--- title: "Agent composability: why AI agents should be built from parts you can swap" url: https://agentcomposability.com/ description: "Why AI agent systems built from swappable parts with open interfaces (models, tools, context, skills) are the right design for this era. Sourced, with the objections." last_checked: 2026-10-21 license: CC BY 4.0 --- # Build AI agents from parts you can swap. Models, tools and standards change every few months. An agent system that treats each of them as a separate part, behind an open interface, keeps working when one part changes. That is agent composability. This page explains why it is the right design for this era, with sources and with the objections. ## What agent composability means **Agent composability** is the ability to build an AI agent system from separate parts (models, tools, skills, sub-agents and other agents) that connect through open, shared interfaces. You can then replace one part without rebuilding the others. This wording combines the three sources below. 1. Amazon Web Services (AWS) Prescriptive Guidance: “In technology, composability is the act of combining modular components to create a flexible, scalable solution that can adapt to change.” — [Focus area 2: Design for composability and collaboration](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-operationalizing-agentic-ai/focus-areas-composability-collaboration.html) 2. Atlan: “Agent composability is the ability to build an AI system out of smaller and independent agents, tools, or resources that can be assembled, swapped, and recombined.” — [What Is Agent Composability?](https://atlan.com/know/ai-agent/agent-composability/) 3. Anthropic: “Consistently, the most successful implementations use simple, composable patterns rather than complex frameworks.” — [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) ## The model is the part that changes fastest. For the same capability, the price of a model falls fast. Epoch AI measured declines “ranging from 9x to 900x per year”, depending on the task, and adds that it is “less clear” that the fastest drops will last. Stanford’s AI Index found that the cost of GPT-3.5-level performance “dropped over 280-fold” between November 2022 and October 2024. Most teams do not leave their vendor. They upgrade. In a mid-2025 survey of 150 builders, 66% upgraded to a newer model from the same vendor and only 11% changed providers. Either way, the model under the system changes, again and again. In a composable system that change is a configuration change. In a monolith it is a rewrite. Sources: [Epoch AI](https://epoch.ai/data-insights/llm-inference-price-trends); [Stanford AI Index](https://hai.stanford.edu/ai-index/2025-ai-index-report); [Menlo Ventures](https://menlovc.com/perspective/2025-mid-year-llm-market-update/). ## An agent is a recipe of parts. Anthropic describes the basic building block of an agent as “an LLM enhanced with augmentations such as retrieval, tools, and memory.” Each augmentation is a part. Give each part one job and one interface, and you can replace it alone. - **Behind a gateway, not hard-coded.** A model gateway puts one endpoint in front of many providers, with fallback, logging and budgets. The application names a model in configuration. Sources: [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/); [Vercel AI Gateway](https://vercel.com/docs/ai-gateway). - **Behind the Model Context Protocol (MCP).** MCP connects any agent to tools and data through one protocol. Stripe runs one internal MCP server, Toolshed, with “more than 400 MCP tools”, and gives each agent a curated subset. Sources: [Stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents). - **In files and stores that you own.** Instructions in AGENTS.md, memory in files or your own database, and retrieval on demand. Anthropic’s advice is to find “the smallest set of high-signal tokens” for each step. Sources: [agents.md compatibility list](https://agents.md/); [Anthropic](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents). - **Packaged procedures that load when needed.** An Agent Skill is a folder with a SKILL.md file. “At startup, the agent pre-loads the name and description of every installed skill”; it reads the rest only when a task needs it. Sources: [Anthropic](https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills); [Agent Skills overview](https://agentskills.io/home). - **Workflows first, agents where they pay.** Workflows follow “predefined code paths”; agents “dynamically direct their own processes and tool usage”. The patterns combine: prompt chaining, routing, parallelization, orchestrator-workers and evaluator-optimizer. Agents can call other agents as tools or hand off a task. Sources: [Anthropic](https://www.anthropic.com/engineering/building-effective-agents); [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/multi_agent/); [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/handoffs/). - **Evaluations, traces and permissions per part.** A test set for each part, a trace for each model call and tool call, and a scoped token for each tool. Under the MCP authorization specification, servers “MUST NOT accept or transit any other tokens”. Sources: [Anthropic](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents); [OpenTelemetry](https://opentelemetry.io/docs/specs/semconv/gen-ai/); [MCP specification](https://modelcontextprotocol.io/specification/2025-06-18/basic/authorization). Sources: [Anthropic](https://www.anthropic.com/engineering/building-effective-agents). ## Open standards make the parts fit. Parts are only swappable when their interfaces are shared. Since 2024 the industry has agreed on a small set: AGENTS.md for instructions, MCP for tools and context, Agent2Agent (A2A) for agents that talk to agents, and Agent Skills for packaged procedures. On 9 December 2025 the Linux Foundation formed the Agentic AI Foundation (AAIF) to host several of them. Anthropic counted “more than 10,000 active public MCP servers” that day. A2A joined the AAIF on 17 August 2026. No single vendor owns the interfaces, so no single vendor owns your system. The standards page has the full table, and a dated matrix of which agents support each standard. Sources: [AAIF formation announcement](https://aaif.io/news/linux-foundation-announces-formation-of-aaif); [Anthropic](https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation); [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif). ## Six reasons composition wins. - **Adopt a better model in days, not quarters.** Anthropic: “Evals also shape how quickly you can adopt new models.” Tests per part plus a model behind a gateway make an upgrade a measured change. Sources: [Anthropic](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents). - **Use the right model for each step.** In 2025, 37% of 100 enterprise chief information officers used five or more models, up from 29% a year earlier. Anthropic’s routing pattern sends easy questions to smaller, cheaper models and hard ones to more capable models. Sources: [a16z](https://a16z.com/ai-enterprise-2025/); [Anthropic](https://www.anthropic.com/engineering/building-effective-agents). - **Keep your context out of any one vendor.** Forrester predicts that “vendor fragmentation will force a majority of enterprises to compose ‘agentlakes.’” Your instructions, procedures and memory are the asset. Kept in open formats, they move with you. Sources: [iTWire (Forrester)](https://itwire.com/it-industry-news/strategy/forrester-unveils-2026-predictions-for-ai-and-tech-leadership.html). - **Give each part only the access it needs.** At Block, “MCP allows us to define which models can invoke which tools, and lets us annotate tools as ‘read-only’ or ‘destructive’ to require user confirmation when necessary.” Sources: [Block](https://goose-docs.ai/blog/2025/04/21/mcp-in-enterprise). - **Build a part once, use it in every agent.** Stripe’s Toolshed serves its coding agents and its other agents from one MCP server. LinkedIn keeps a central skill registry, where “skills encapsulate both what and how”. Sources: [Stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents); [InfoQ (LinkedIn)](https://www.infoq.com/presentations/LinkedIn-agent-hiring-assistant/). - **Ride the ecosystem instead of building it.** Every MCP server, skill and A2A agent that someone else builds is a part that your system can use. GitHub’s Agent HQ runs coding agents from several vendors in one place, with an MCP registry and custom agents through AGENTS.md. Sources: [GitHub](https://github.blog/news-insights/company-news/welcome-home-agents/). ## This idea has won before. Every era of computing met the same problem and chose the same answer: small parts, one shared interface, and the freedom to replace a part without touching the rest. Agents are the latest case. > “Expect the output of every program to become the input to another, as yet unknown, program.” — M. D. McIlroy, E. N. Pinson and B. A. Tague, Bell System Technical Journal, 1978 - **Hide what will change behind an interface.** Parnas: begin with “design decisions which are likely to change. Each module is then designed to hide such a decision from the others.” Today the decision most likely to change is the model. Sources: [Parnas 1972](https://doi.org/10.1145/361598.361623). - **Small programs, one universal interface.** “Make each program do one thing well.” Text streams joined the programs. For agents, tool schemas, MCP messages and SKILL.md files do the same job. Sources: [Bell System Technical Journal](https://danluu.com/mcilroy-unix/). - **A uniform interface lets each side evolve alone.** Fielding on REST: “Implementations are decoupled from the services they provide, which encourages independent evolvability.” He also names the cost: “a uniform interface degrades efficiency.” Sources: [Fielding 2000](https://ics.uci.edu/~fielding/pubs/dissertation/rest_arch_style.htm). - **Components you can replace one at a time.** Lewis and Fowler define a component as “a unit of software that is independently replaceable and upgradeable.” Sources: [Lewis and Fowler](https://martinfowler.com/articles/microservices.html). - **From one model to compound systems.** Berkeley researchers: “state-of-the-art AI results are increasingly obtained by compound systems with multiple components, not just monolithic models.” Agent standards turn those components into parts that anyone can build and swap. Sources: [BAIR](https://bair.berkeley.edu/blog/2024/02/18/compound-ai-systems/). ## Teams that already build this way. Each case below is a company describing its own system (own claims). Each one composes separate parts behind shared interfaces. - **Coding agents on a shared tool server.** Minions run on a fork of goose, reach “more than 400 MCP tools” through Toolshed, mix agent loops with fixed steps (git, linters, tests) and produce “more than a thousand pull requests merged each week”. Sources: [Stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents). - **One agent, many models, many MCP servers.** goose connects models from more than one vendor to MCP servers for Snowflake, GitHub, Jira, Slack, Google Drive and internal systems, with per-tool permissions. Sources: [Block](https://goose-docs.ai/blog/2025/04/21/mcp-in-enterprise). - **A supervisor agent and a skill registry.** The Hiring Assistant uses a supervisor that calls sub-agents as tools, a central skill registry, an inference layer that changes models by configuration, and a grading rubric per component. Sources: [InfoQ (LinkedIn)](https://www.infoq.com/presentations/LinkedIn-agent-hiring-assistant/). - **Few tools per agent, judges calibrated against people.** Sidekick’s team found that with “50+ tools” a system becomes “difficult to reason about”, and its advice is to “Avoid multi-agent architectures early”. Composable, but simple. Sources: [Shopify](https://shopify.engineering/building-production-ready-agentic-systems). - **A cloud sells the parts separately.** Amazon Bedrock AgentCore offers runtime, memory, gateway, identity and observability as “composable AgentCore services that can be used together or independently”, with any model or framework. Sources: [AWS](https://aws.amazon.com/blogs/machine-learning/amazon-bedrock-agentcore-is-now-generally-available/). ## Composable does not mean many chatty agents. The strongest critics of multi-agent systems are right about what they criticize. Their point is that agents which talk freely to each other fail. A composable system is a different design: one coordinator owns the decisions and the writes, and the parts behind it have narrow jobs and narrow interfaces. Here are the objections, each with its source, and the design answer to each one. - **“Running multiple agents in collaboration only results in fragile systems.”** Cognition, June 2025. In April 2026 it reported what does work: setups “where agents contribute intelligence while writes stay single-threaded.” Answer: one coordinator writes; the parts read, check and advise. Sources: [Cognition](https://cognition.ai/blog/dont-build-multi-agents); [Cognition](https://cognition.ai/blog/multi-agents-working). - **Multi-agent systems fail in 14 known ways.** The MAST study found “14 unique modes, clustered into 3 categories”: system design issues, inter-agent misalignment and task verification. Answer: typed interfaces between parts, and an explicit verification step. Sources: [MAST (arXiv 2503.13657)](https://arxiv.org/abs/2503.13657). - **Independent agents multiply errors.** Google Research found that “independent multi-agent systems amplify errors by up to 17.2x”, and that on sequential planning “every multi-agent variant we tested degraded performance by 39-70%.” Answer: a central coordinator, and no split of tasks that are sequential by nature. Sources: [Google Research](https://research.google/blog/towards-a-science-of-scaling-agent-systems-when-and-why-agent-systems-work/). - **Multi-agent systems burn tokens.** Anthropic’s research system used about 15 times the tokens of a chat, and scored 90.2% higher than a single agent on Anthropic’s own research evaluation. Answer: split only wide, read-heavy work; send small steps to small models. Sources: [Anthropic](https://www.anthropic.com/engineering/multi-agent-research-system); [Microsoft Learn](https://learn.microsoft.com/en-us/azure/architecture/ai-ml/guide/ai-agent-design-patterns). - **Every new part is a new door.** In tool poisoning, “malicious instructions are embedded within MCP tool descriptions that are invisible to users but visible to AI models.” Simon Willison warns that MCP “encourages users to mix and match tools from different sources”, and names a lethal trifecta: private data, untrusted content and external communication. This is the strongest objection. Answer: pinned versions, scoped tokens per part, and no single agent that holds all three. Sources: [Invariant Labs](https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks); [Simon Willison](https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/). - **Most agent projects may not survive.** Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027. Answer: small parts with a test and a cost each show early which part fails, before the whole project does. Sources: [W.Media (Gartner)](https://w.media/over-40-percent-of-agentic-ai-projects-could-face-the-axe-by-end-of-2027-gartner/). - **Teams rarely switch, and agents make switching harder.** a16z found that “the rise of agentic workflows has started making it more difficult to switch between models.” That is the case for composability, not against it: the cost of a switch comes from tight coupling. Sources: [a16z](https://a16z.com/ai-enterprise-2025/). - **Better models will make scaffolding obsolete.** Sutton’s bitter lesson: “general methods that leverage computation are ultimately the most effective, and by a large margin.” Answer: keep the model layer swappable and the scaffolding thin, so that each better model makes the system better without a rebuild. Sources: [Sutton](http://www.incompleteideas.net/IncIdeas/BitterLesson.html). ## Ten checks for a composable agent system. - **A gateway sits between your code and the models.** A model change is a configuration change, with fallback. Sources: [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/). - **Tools sit behind MCP servers.** Each agent gets a curated subset, not every tool. Sources: [Stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents). - **Each agent has few, clearly named tools.** Sources: [Shopify](https://shopify.engineering/building-production-ready-agentic-systems); [Anthropic](https://www.anthropic.com/engineering/writing-tools-for-agents). - **Context loads just in time.** Skill names first, full procedures and memory only when needed. Sources: [Anthropic](https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills); [Anthropic](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents). - **Instructions and memory live in files you own.** Sources: [agents.md compatibility list](https://agents.md/). - **Workflows first; sub-agents only where a task splits cleanly.** Sources: [Anthropic](https://www.anthropic.com/engineering/building-effective-agents); [Shopify](https://shopify.engineering/building-production-ready-agentic-systems). - **Fixed code steps wrap the agent loops.** Tests, linters and checks run as code, not as prompts. Sources: [Stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents). - **Every part has an evaluation before you swap it.** Sources: [Anthropic](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents). - **Every model call and tool call leaves a trace.** Sources: [OpenTelemetry](https://opentelemetry.io/docs/specs/semconv/gen-ai/). - **Every agent has its own identity and scoped access.** Destructive tools need a human approval. Sources: [MCP specification](https://modelcontextprotocol.io/specification/2025-06-18/basic/authorization); [Block](https://goose-docs.ai/blog/2025/04/21/mcp-in-enterprise). ## Questions ### What is agent composability? Agent composability is the ability to build an AI agent system from separate parts that connect through open, shared interfaces. You can replace one part without rebuilding the others. Amazon Web Services describes composability as “the act of combining modular components to create a flexible, scalable solution that can adapt to change.” Sources: [AWS Prescriptive Guidance](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-operationalizing-agentic-ai/focus-areas-composability-collaboration.html); [Atlan](https://atlan.com/know/ai-agent/agent-composability/). ### Why is composability the right design for AI agents now? Because the parts change faster than the systems. Epoch AI measured model price declines “ranging from 9x to 900x per year”, depending on the task, and most teams upgrade their model within months. A system built from parts with open interfaces takes each new model, tool or standard as a configuration change instead of a rewrite. Sources: [Epoch AI](https://epoch.ai/data-insights/llm-inference-price-trends); [Menlo Ventures](https://menlovc.com/perspective/2025-mid-year-llm-market-update/). ### Does a composable agent system mean many agents? No. A composable system can have one agent. What makes it composable is that its model, tools, context and skills are separate parts behind shared interfaces. When a system does use several agents, the designs that work keep one coordinator that owns the writes: Cognition reports that what works is “where agents contribute intelligence while writes stay single-threaded.” Sources: [Cognition](https://cognition.ai/blog/multi-agents-working); [Anthropic](https://www.anthropic.com/engineering/building-effective-agents). ### What does “composable agents” mean? Composable agents are agents that you build as independent units, so that you can combine, swap and reuse them. Amazon Web Services states the goal like this: “Let me build agents like I build services – modular and testable, so that they can be composed and orchestrated as needed.” In practice, an orchestrator agent can split a task and send the parts to specialized sub-agents. Sources: [AWS Prescriptive Guidance](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-operationalizing-agentic-ai/focus-areas-composability-collaboration.html); [Atlan](https://atlan.com/know/ai-agent/agent-composability/). ### Is composability the same as interoperability? No, but the two are linked. Interoperability lets independent agents discover one another, communicate and exchange work across systems. Composability is the design property that uses this: you build a system from parts that you can combine and replace. Sources: [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif); [Atlan](https://atlan.com/know/ai-agent/agent-composability/); [AWS Prescriptive Guidance](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-operationalizing-agentic-ai/focus-areas-composability-collaboration.html). ### How do MCP and A2A differ? MCP connects an agent to tools, data sources, applications and services. A2A connects agents to other agents, so that they can discover one another, delegate tasks and exchange results. The A2A project calls the two protocols complementary: “A2A connects the agents to each other; MCP connects each agent to its own tools.” Sources: [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif); [A2A Protocol: A2A and MCP](https://a2a-protocol.org/latest/topics/a2a-and-mcp/). ### What are Agent Skills? Agent Skills are “a lightweight, open format for extending AI agent capabilities with specialized knowledge and workflows.” A skill is a folder with a SKILL.md file that holds metadata and instructions, and it can also hold scripts and other resources. At startup, an agent loads only the name and description of each skill, and it reads the full instructions when a task needs them. Sources: [Agent Skills overview](https://agentskills.io/home). ## Next - [The standards and the compatibility matrix](https://agentcomposability.com/standards) - [Work with us: we design composable agent systems for US companies](https://agentcomposability.com/work) - [This domain is for sale](https://agentcomposability.com/domain) --- title: "Agent composability standards: which AI agents support MCP, A2A, Agent Skills and AGENTS.md (2026)" url: https://agentcomposability.com/standards description: "The open standards layer for composable AI agents, and a dated matrix of which agents support Agent Skills, AGENTS.md, MCP extensions and A2A. Every cell cites an official source." last_checked: 2026-10-21 license: CC BY 4.0 --- # The standards, and who supports them Composable parts need shared interfaces. This page lists the open standards for each layer of an agent system, and a dated matrix of which agents an official source lists for each standard. ## The open standards that make agents composable Composable parts need shared interfaces. The Agentic AI Foundation (AAIF) names five layers of an “open agentic stack”, and one of its projects covers each layer. The Linux Foundation hosts the AAIF. The table adds Agent Skills, two MCP extensions, and projects for discovery and conformance testing. ### The five AAIF layers | Layer | Standard | Steward and status | Sources | |---|---|---|---| | Instructions and context | AGENTS.md | OpenAI released it in August 2025. OpenAI gave it to the AAIF when the AAIF formed on 9 December 2025. | [AAIF formation announcement](https://aaif.io/news/linux-foundation-announces-formation-of-aaif); [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif) | | Agent runtime | goose | Block gave it to the AAIF on 9 December 2025. | [AAIF formation announcement](https://aaif.io/news/linux-foundation-announces-formation-of-aaif); [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif) | | Agent to tool connectivity | Model Context Protocol (MCP) | Anthropic released it as open source in November 2024. Anthropic gave it to the AAIF on 9 December 2025. | [AAIF formation announcement](https://aaif.io/news/linux-foundation-announces-formation-of-aaif); [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif) | | Traffic mediation and control | agentgateway | An AAIF project. It sits between agent systems and their infrastructure, and it manages routing, policy and observability. | [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif) | | Agent to agent interoperability | Agent2Agent (A2A) | Google launched it in April 2025 and gave it to the Linux Foundation. It joined the AAIF as a hosted project on 17 August 2026. More than 150 organizations back it. | [AAIF: A2A joins AAIF](https://aaif.io/blog/a2a-joins-aaif) | ### Skills and MCP extensions | Layer | Standard | Steward and status | Sources | |---|---|---|---| | Packaged procedures | Agent Skills | A skill is a folder with a SKILL.md file. Anthropic first developed the format and released it as an open standard. | [Agent Skills overview](https://agentskills.io/home) | | Skills inside MCP | Skills over MCP extension (`io.modelcontextprotocol/skills`) | The specification has Final status. The Skills Over MCP Working Group maintains it. Hosts and software development kits are still adding support. | [MCP Skills extension](https://modelcontextprotocol.io/extensions/skills/overview) | | Interactive interfaces inside MCP | MCP Apps extension (`io.modelcontextprotocol/ui`) | An extension to the core MCP specification. MCP servers can return interactive interfaces, such as forms and dashboards, that show in the chat. Host support varies by client. | [MCP Apps overview](https://modelcontextprotocol.io/extensions/apps/overview) | ### Discovery | Layer | Standard | Steward and status | Sources | |---|---|---|---| | Find MCP servers | Official MCP Registry | In preview. The official central metadata repository for publicly accessible MCP servers. | [The MCP Registry](https://modelcontextprotocol.io/registry/about) | | Find MCP servers | GitHub MCP Registry | GitHub launched it on 16 September 2025 as a place to discover MCP servers. | [The GitHub Blog](https://github.blog/ai-and-ml/github-copilot/meet-the-github-mcp-registry-the-fastest-way-to-discover-mcp-servers/) | | Find A2A agents and MCP servers | AGNTCY directories | Cisco released AGNTCY as open source in March 2025. The Linux Foundation welcomed the project on 29 July 2025. Its directories make A2A agents and MCP servers discoverable. | [Linux Foundation press release](https://www.linuxfoundation.org/press/linux-foundation-welcomes-the-agntcy-project-to-standardize-open-multi-agent-system-infrastructure-and-break-down-ai-agent-silos) | ### Conformance testing | Layer | Standard | Steward and status | Sources | |---|---|---|---| | Test MCP clients and servers | MCP Conformance Test Framework | Tests MCP client and server implementations against the specification. | [modelcontextprotocol/conformance](https://github.com/modelcontextprotocol/conformance) | | Test A2A implementations | A2A Technology Compatibility Kit (TCK) | A compatibility test suite for A2A implementations on three transports. | [a2aproject/a2a-tck](https://github.com/a2aproject/a2a-tck) | ## Which agents support which standards The matrix shows which agents each official source lists. The matrix covers six agents: Claude Code, ChatGPT and Codex, Cursor, GitHub Copilot with Visual Studio Code (VS Code), Gemini CLI (command-line interface) and goose. Each column names its source and the date of the last check. In the A2A column, each cell cites the documentation of that client. - **Listed**: An official source names the client as supporting the standard. - **Listed, partial**: The source names the client and labels its support “Partial”. - **Not listed**: The source that we checked does not name the client. This makes no claim either way. | Client | Agent Skills | AGENTS.md | MCP Apps | Skills over MCP | A2A | |---|---|---|---|---|---| | Claude Code | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://code.claude.com/docs/en/memory#agents-md) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://code.claude.com/docs/en/overview) (2026-10-21) | | ChatGPT and Codex | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://agents.md/) (2026-10-21) | [Listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Listed, partial](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://learn.chatgpt.com/docs) (2026-10-21) | | Cursor | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://agents.md/) (2026-10-21) | [Listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://cursor.com/docs) (2026-10-21) | | GitHub Copilot and VS Code | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://agents.md/) (2026-10-21) | [Listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://docs.github.com/en/copilot) (2026-10-07) | | Gemini CLI | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://agents.md/) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Listed](https://geminicli.com/docs/core/remote-agents/) (2026-10-21) | | goose | [Listed](https://agentskills.io/clients) (2026-10-21) | [Listed](https://agents.md/) (2026-10-21) | [Listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://modelcontextprotocol.io/extensions/client-matrix) (2026-10-21) | [Not listed](https://goose-docs.ai/docs/quickstart/) (2026-10-07) | Data with the source and check date of every cell: [matrix.json](https://agentcomposability.com/matrix.json). ## Composability, orchestration and chaining People often use these three words as if they mean the same thing. They do not. Composability is a design property. Orchestration and chaining are runtime patterns: they describe how the parts run. - **Composability** (Design property): You build the system from independent units. You can swap or recombine the units. Sources: [Atlan](https://atlan.com/know/ai-agent/agent-composability/). - **Orchestration** (Runtime pattern): A central agent breaks a task into parts while it runs. It gives the parts to worker agents and combines their results. It suits tasks with subtasks that you cannot predict. Sources: [Atlan](https://atlan.com/know/ai-agent/agent-composability/); [Anthropic](https://www.anthropic.com/engineering/building-effective-agents). - **Chaining** (Runtime pattern): The steps run in a fixed sequence. The output of each step is the input of the next step. It suits workflows that you can predict. Sources: [Atlan](https://atlan.com/know/ai-agent/agent-composability/); [Anthropic](https://www.anthropic.com/engineering/building-effective-agents). Composability makes both runtime patterns possible. Atlan gives an example: units with clean boundaries “can be chained today and orchestrated tomorrow without a rewrite.” ## How this page stays true - We check each source by hand. - Each matrix cell uses one official source: the list that the standard publishes, or the documentation of the client. - Each cell records its source and the date of its last check. When a source changes, we change the cell and its date. --- title: "Design your company’s agent system | agentcomposability.com" url: https://agentcomposability.com/work description: "We design composable AI agent systems for US companies: models you can swap, tools behind open standards, context you own, and permissions you can audit." last_checked: 2026-10-21 license: CC BY 4.0 --- # We design agent systems that outlive the next model. The model you pick today will be replaced within a year. We design the system around it so that the replacement is a configuration change, not a rebuild. Remote, for companies in the United States. ## Three steps, each one useful on its own - **Map the work** (Step 01) We sit with the people who do the work. We list the tasks, the systems they touch and the decisions they make. We mark where an agent pays for itself and where it must not act alone. - **Design the system** (Step 02) We draw the parts and the interfaces: which model does which step, which tools connect through the Model Context Protocol (MCP), where your company’s context lives, which skills the agents load, who approves what, and how you measure each part. - **Build the first agent and hand it over** (Step 03) We build the first working agent on that design, with your team, inside your tools. You get the code, the runbooks and the evaluations. Your team owns and runs it after we leave. ## Four rules we do not trade - **Any model can be replaced.** (Swap) The design names an interface, not a vendor. A cheaper or better model goes in without a rewrite. - **Your context stays yours.** (Own) Instructions, procedures and memory live in plain files and in your own stores, not inside one vendor’s product. - **Each part gets only the access it needs.** (Limit) Every tool connection has its own permission. A human approves the actions that cannot be undone. - **Each part has a test.** (Measure) We write evaluations per part, so that you can change one part and know what changed. ## Who this is for Founders, operators and engineering leaders at US companies who want agents in real work, not in a demo. We work as forward deployed engineers: in your tools, with your team, on your problem. ## Contact Describe the work you want agents to do. We reply within two business days. Use the form at https://agentcomposability.com/work#contact. An agent may send the form for a person who asked it to: POST https://agentcomposability.com/api/contact as application/x-www-form-urlencoded with the fields intent=design, name, email, company and message. --- title: "For AI agents: llms.txt, Markdown, JSON and an MCP server | agentcomposability.com" url: https://agentcomposability.com/agents description: "How AI agents can read agentcomposability.com: llms.txt, Markdown copies of every page, the compatibility matrix as JSON, and a read-only Model Context Protocol server." last_checked: 2026-10-21 license: CC BY 4.0 --- # This site is built the way it argues. Every page is a part with an open interface. People read the HTML. Agents read the same facts as Markdown, as JSON or through a Model Context Protocol (MCP) server. All of it is read-only, free and needs no key. ## Interfaces - **/llms.txt** (Summary): A curated summary of the site and the matrix (llmstxt.org format). - **/llms-full.txt** (Full text): Every page in one Markdown file. - **Accept: text/markdown** (Markdown): Every page has a Markdown copy at the same path plus .md (/index.md for the home page). Send the header Accept: text/markdown to any page URL and you get the Markdown copy. - **/matrix.json** (Data): The compatibility matrix: one entry per client, with the status, source URL and check date of each cell. CORS is open. - **/mcp** (MCP): A read-only MCP server over Streamable HTTP. No sign-in. Four tools, listed below. - **/robots.txt and /sitemap.xml** (Crawl): AI crawlers and search crawlers are welcome. Content signals: search, AI input and AI training are all allowed. ## MCP tools - `define_agent_composability`: The sourced definition, with its quotes and URLs. - `get_compatibility_matrix`: Which agents an official source lists for each standard. Optional filters: client and standard. - `list_standards`: The standards layer: AGENTS.md, MCP, A2A, Agent Skills and the rest, with status and sources. - `read_page`: Any page of this site as Markdown: why, standards, work, agents, domain, sources. ## Connect Terminal · Claude Code: ``` claude mcp add --transport http agentcomposability https://agentcomposability.com/mcp ``` JSON · MCP client configuration: ``` { "mcpServers": { "agentcomposability": { "type": "http", "url": "https://agentcomposability.com/mcp" } } } ``` Terminal · Test with curl: ``` curl -s https://agentcomposability.com/mcp \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' ``` ## Rules for citing - Cite the source URL of a matrix cell, not only this site. Each cell names its source and check date. - “Not listed” means that the source we checked does not name the client. It is not proof that the client has no support. - For the newest state, follow the source link. Our check dates are in matrix.json. --- title: "agentcomposability.com is for sale | agentcomposability.com" url: https://agentcomposability.com/domain description: "The exact-match .com for agent composability is for sale, with or without the live reference site, its sourced data and its Model Context Protocol server." last_checked: 2026-10-21 license: CC BY 4.0 --- # agentcomposability.com is for sale. The exact-match .com domain for agent composability: the design idea behind open agent standards. You can buy the domain alone, or the domain with this live site. ## A name for an idea that is growing - **An exact-match .com for a rising term.** (The name) Agent papers on arXiv that use “composable” or “composability” went from 0.19% in 2025 to 0.53% in 2026. Google suggests “agent composability” as a search. No standards body owns the term yet. - **A live, sourced reference.** (The site) A definition, the standards layer and a dated compatibility matrix in which every cell cites an official source. Optional in the sale. - **Ready for AI agents.** (For agents) llms.txt, a Markdown copy of each page, an open JSON dataset and a read-only MCP server. Optional in the sale. - **Who it suits.** (Fit) An agent platform, an interoperability standard or foundation, an integration company, or a publication about AI agents. The arXiv figures come from a study of 2026-10-07 that counted agent papers on the arXiv API. The Google suggestion was seen the same day. ## How a sale works 1. Send an offer with the form below. 2. We reply within two business days. 3. Payment and transfer go through Escrow.com, or another escrow service that you choose. The buyer pays the escrow fee. ## Make an offer Offers in US dollars. Tell us if you want the domain alone or the domain with the site. Use the form at https://agentcomposability.com/domain#contact, or POST https://agentcomposability.com/api/contact as application/x-www-form-urlencoded with intent=domain, name, email, company, offer and message. --- title: "Sources | agentcomposability.com" url: https://agentcomposability.com/sources description: "Every source that agentcomposability.com cites." last_checked: 2026-10-21 license: CC BY 4.0 --- # Every source, with its date - A2A project on GitHub: [A2A Protocol Technology Compatibility Kit (TCK)](https://github.com/a2aproject/a2a-tck). Checked 2026-10-07. - A2A Protocol: [A2A and MCP](https://a2a-protocol.org/latest/topics/a2a-and-mcp/). Checked 2026-10-07. - Agent Skills: [Agent Skills Overview](https://agentskills.io/home). Checked 2026-10-07. - Agent Skills: [Client Showcase](https://agentskills.io/clients). Checked 2026-10-07. - Agentic AI Foundation (AAIF): [Linux Foundation Announces the Formation of the Agentic AI Foundation (AAIF)](https://aaif.io/news/linux-foundation-announces-formation-of-aaif). Published 2025-12-09, checked 2026-10-07. - Agentic AI Foundation (AAIF): [A2A joins AAIF’s open agentic stack](https://aaif.io/blog/a2a-joins-aaif). Published 2026-08-17, checked 2026-10-07. - AGENTS.md: [One AGENTS.md works across many agents](https://agents.md/). Checked 2026-10-07. - Amazon Web Services (AWS): [Make agents a reality with Amazon Bedrock AgentCore: now generally available](https://aws.amazon.com/blogs/machine-learning/amazon-bedrock-agentcore-is-now-generally-available/). Published 2025-10-13, checked 2026-10-21. - Amazon Web Services (AWS) Prescriptive Guidance: [Focus area 2: Design for composability and collaboration](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-operationalizing-agentic-ai/focus-areas-composability-collaboration.html). No date on the page, checked 2026-10-07. - Andreessen Horowitz (a16z): [How 100 enterprise CIOs are building and buying gen AI in 2025](https://a16z.com/ai-enterprise-2025/). Published 2025-06-10, checked 2026-10-21. - Anthropic: [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents). Published 2024-12-19, checked 2026-10-07. - Anthropic: [Equipping agents for the real world with Agent Skills](https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills). Published 2025-10-16, checked 2026-10-21. - Anthropic: [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents). Published 2025-09-29, checked 2026-10-21. - Anthropic: [Demystifying evals for AI agents](https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents). Published 2026-01-09, checked 2026-10-21. - Anthropic: [How we built our multi-agent research system](https://www.anthropic.com/engineering/multi-agent-research-system). Published 2025-06-13, checked 2026-10-21. - Anthropic: [Writing effective tools for AI agents](https://www.anthropic.com/engineering/writing-tools-for-agents). Published 2025-09-11, checked 2026-10-21. - Anthropic: [Donating MCP to the Agentic AI Foundation](https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation). Published 2025-12-09, checked 2026-10-21. - arXiv: [Cemri and others, “Why do multi-agent LLM systems fail?”](https://arxiv.org/abs/2503.13657). Published 2025-03-17, checked 2026-10-21. - Atlan: [What Is Agent Composability?](https://atlan.com/know/ai-agent/agent-composability/). Published 2026-09-11, updated 2026-09-21, checked 2026-10-07. - Bell System Technical Journal (transcription by Dan Luu): [M. D. McIlroy, E. N. Pinson and B. A. Tague, “UNIX Time-Sharing System: Foreword”, volume 57, number 6](https://danluu.com/mcilroy-unix/). Published July–August 1978, checked 2026-10-21. - Berkeley Artificial Intelligence Research (BAIR) blog: [The shift from models to compound AI systems](https://bair.berkeley.edu/blog/2024/02/18/compound-ai-systems/). Published 2024-02-18, checked 2026-10-21. - ChatGPT Learn: [ChatGPT and Codex documentation](https://learn.chatgpt.com/docs). Checked 2026-10-07. - Claude Code Docs: [How Claude remembers your project: AGENTS.md](https://code.claude.com/docs/en/memory#agents-md). Checked 2026-10-07. - Claude Code Docs: [Claude Code documentation](https://code.claude.com/docs/en/overview). Checked 2026-10-07. - Cloudflare: [AI Gateway overview](https://developers.cloudflare.com/ai-gateway/). Updated 2026-04-20, checked 2026-10-21. - Cognition: [Don’t build multi-agents](https://cognition.ai/blog/dont-build-multi-agents). Published 2025-06-12, checked 2026-10-21. - Cognition: [Multi-agents: what’s actually working](https://cognition.ai/blog/multi-agents-working). Published 2026-04-22, checked 2026-10-21. - Communications of the ACM: [D. L. Parnas, “On the criteria to be used in decomposing systems into modules”, volume 15, number 12](https://doi.org/10.1145/361598.361623). Published December 1972, checked 2026-10-21. - Cursor Docs: [Cursor documentation](https://cursor.com/docs). Checked 2026-10-07. - Epoch AI: [LLM inference prices have fallen rapidly but unequally across tasks](https://epoch.ai/data-insights/llm-inference-price-trends). Published 2025-03-12, checked 2026-10-21. - Gemini CLI documentation: [Remote Subagents](https://geminicli.com/docs/core/remote-agents/). Checked 2026-10-07. - GitHub Docs: [GitHub Copilot documentation](https://docs.github.com/en/copilot). Checked 2026-10-07. - Google Research: [Towards a science of scaling agent systems: when and why agent systems work](https://research.google/blog/towards-a-science-of-scaling-agent-systems-when-and-why-agent-systems-work/). Published 2026-01-28, checked 2026-10-21. - goose blog (Block): [MCP in the enterprise: real world adoption at Block](https://goose-docs.ai/blog/2025/04/21/mcp-in-enterprise). Published 2025-04-21, checked 2026-10-21. - goose documentation: [goose documentation](https://goose-docs.ai/docs/quickstart/). Checked 2026-10-07. - InfoQ (talk by LinkedIn engineers): [Lessons learned from building LinkedIn’s first agent: Hiring Assistant](https://www.infoq.com/presentations/LinkedIn-agent-hiring-assistant/). Published 2025-12-31, checked 2026-10-21. - Invariant Labs: [MCP security notification: tool poisoning attacks](https://invariantlabs.ai/blog/mcp-security-notification-tool-poisoning-attacks). Published 2025-04-01, checked 2026-10-21. - iTWire, reporting Forrester's 2026 predictions: [Forrester 2026 predictions for AI and tech leadership](https://itwire.com/it-industry-news/strategy/forrester-unveils-2026-predictions-for-ai-and-tech-leadership.html). Published 2025-10-10, checked 2026-10-21. - martinfowler.com: [James Lewis and Martin Fowler, “Microservices”](https://martinfowler.com/articles/microservices.html). Published 2014-03-25, checked 2026-10-21. - Menlo Ventures: [2025 mid-year LLM market update: foundation model landscape and economics](https://menlovc.com/perspective/2025-mid-year-llm-market-update/). Published 2025-07-31, checked 2026-10-21. - Microsoft Learn (Azure Architecture Center): [AI agent orchestration patterns](https://learn.microsoft.com/en-us/azure/architecture/ai-ml/guide/ai-agent-design-patterns). Updated 2026-09-21, checked 2026-10-21. - Model Context Protocol: [Extension Support Matrix](https://modelcontextprotocol.io/extensions/client-matrix). Checked 2026-10-07. - Model Context Protocol: [MCP Apps](https://modelcontextprotocol.io/extensions/apps/overview). Checked 2026-10-07. - Model Context Protocol: [Skills](https://modelcontextprotocol.io/extensions/skills/overview). Checked 2026-10-07. - Model Context Protocol: [The MCP Registry](https://modelcontextprotocol.io/registry/about). Checked 2026-10-07. - Model Context Protocol: [Authorization (specification version 2025-06-18)](https://modelcontextprotocol.io/specification/2025-06-18/basic/authorization). Published 2025-06-18, checked 2026-10-21. - Model Context Protocol on GitHub: [MCP Conformance Test Framework](https://github.com/modelcontextprotocol/conformance). Checked 2026-10-07. - OpenAI Agents SDK documentation: [Agent orchestration](https://openai.github.io/openai-agents-python/multi_agent/). No date on the page, checked 2026-10-21. - OpenAI Agents SDK documentation: [Handoffs](https://openai.github.io/openai-agents-python/handoffs/). No date on the page, checked 2026-10-21. - OpenAI Developers: [OpenAI developer documentation](https://developers.openai.com/). Checked 2026-10-07. - OpenTelemetry: [Generative AI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/). No date on the page, checked 2026-10-21. - Rich Sutton: [The bitter lesson](http://www.incompleteideas.net/IncIdeas/BitterLesson.html). Published 2019-03-13, checked 2026-10-21. - Shopify Engineering: [Building production-ready agentic systems: lessons from Shopify Sidekick](https://shopify.engineering/building-production-ready-agentic-systems). Published 2025-08-26, checked 2026-10-21. - Simon Willison: [The lethal trifecta for AI agents: private data, untrusted content, and external communication](https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/). Published 2025-06-16, checked 2026-10-21. - Stanford Institute for Human-Centered AI (HAI): [The 2025 AI Index Report](https://hai.stanford.edu/ai-index/2025-ai-index-report). Published April 2025, checked 2026-10-21. - Stripe: [Minions: Stripe's one-shot, end-to-end coding agents](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents). Published 2026-02-09, checked 2026-10-21. - The GitHub Blog: [Meet the GitHub MCP Registry: The fastest way to discover MCP Servers](https://github.blog/ai-and-ml/github-copilot/meet-the-github-mcp-registry-the-fastest-way-to-discover-mcp-servers/). Published 2025-09-16, updated 2025-09-18, checked 2026-10-07. - The GitHub Blog: [Introducing Agent HQ: any agent, any way you work](https://github.blog/news-insights/company-news/welcome-home-agents/). Published 2025-10-28, checked 2026-10-21. - The Linux Foundation: [Linux Foundation Welcomes the AGNTCY Project to Standardize Open Multi-Agent System Infrastructure and Break Down AI Agent Silos](https://www.linuxfoundation.org/press/linux-foundation-welcomes-the-agntcy-project-to-standardize-open-multi-agent-system-infrastructure-and-break-down-ai-agent-silos). Published 2025-07-29, checked 2026-10-07. - University of California, Irvine: [Roy T. Fielding, dissertation, chapter 5: Representational State Transfer (REST)](https://ics.uci.edu/~fielding/pubs/dissertation/rest_arch_style.htm). Published 2000, checked 2026-10-21. - Vercel: [Vercel AI Gateway: models, routing, and observability](https://vercel.com/docs/ai-gateway). Updated 2026-09-14, checked 2026-10-21. - Visual Studio Code documentation: [Build with AI in VS Code](https://code.visualstudio.com/docs/agents/overview). Checked 2026-10-07. - W.Media, reporting a Gartner press release: [Over 40 percent of agentic AI projects could face the axe by end of 2027: Gartner](https://w.media/over-40-percent-of-agentic-ai-projects-could-face-the-axe-by-end-of-2027-gartner/). Published 2025-06-25, checked 2026-10-21.