faf-trinity
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@faf-trinityremember that the production endpoint is https://api.faf.one"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
faf-trinity
faf-trinity — the chassis for context, memory & agent in MCP
This reference implementation shows an MCP server exposing all three IANA-registered FAF formats — .faf (context), .fafm (memory), .fafa (agent identity) — through two mechanisms already proven live in production, not invented for this repo.
Problem
MCP servers today have no standard way to answer "who am I, what do I remember, what's my project." Every server that wants this invents its own ad-hoc shape.
Related MCP server: A-Modular-Kingdom
Solution
The same three files back both exposure mechanisms:
project.faf ──┐
project.fafm ─┼──► Server Card _meta (one.faf/context, one.faf/memory, one.faf/agent)
.fafa ────────┘ └► ai-catalog.json (3 sibling entries, same 3 files)Neither mechanism is new protocol surface — both are already running in production at faf.one and context.faf.one, months before this repo existed. This extracts the minimal, forkable pattern from a full production app down to something you can read in one sitting.
What's actually proven, not just described
npm run demo runs all three, live:
Context — the same BEFORE/AFTER pattern as mcp-project-context: a plain
callTool()hits a "required, none supplied" wall, then the same call routed throughcallToolWithContext()gets filled fromproject.faf.Memory —
remember()a fact on one server process, kill that process entirely, spawn a fresh one,recall()the same fact. No in-memory state survives that — only the file does. That's the actual claim.fafmmakes ("memory that survives across sessions"), proven by genuinely crossing a process boundary, not simulated.Identity —
whoami()reads this server's own.well-known/fafa, and the same three files are shown as a Server Card_metablock — the second proven exposure mechanism.
Render project.faf
npx faf-cli showRenders project.faf as an HTML card. This is what faf show does today — .faf only. .fafm/.fafa rendering isn't part of faf-cli yet; if that ever ships, it's a bonus, not something this repo is waiting on.
Demo
npm install && npm run demoGenerate the catalog
npm run catalogWrites .well-known/ai-catalog.json from the same three source files (project.faf, project.fafm, .well-known/fafa). Generated, gitignored — run it yourself rather than trust a committed copy that could go stale.
Core logic
src/context.ts— the param-fill mechanism, generalized frommcp-project-context(no domain stub this time)src/memory.ts— real, file-backed remember/recall againstproject.fafmitselfsrc/identity.ts—whoami()+ the Server Card_metatrinity blocksrc/catalog-gen.ts— generates theai-catalog.jsonsibling entriessrc/server.ts— wires all three into one MCP server
What this is
A reference implementation of context + memory + agent identity, together, for discussion.
What this is not
Not a library or package intended for installation
Not an npm package — GitHub template distribution, fork it and own it
Not a framework — no plugin system, no API-stability promise
Not a full project-context system
Related
mcp-project-context — the direct predecessor, one format (context) instead of three
application/vnd.faf+yaml,application/vnd.fafm+yaml,application/vnd.fafa+yaml— IANA media type registrations
MIT
This server cannot be installed
Maintenance
Related MCP Servers
- AlicenseBqualityCmaintenanceA local-first MCP server that provides AI agents with safe codebase access through file discovery, hybrid lexical-semantic search, and project introspection. It features durable local memory and semantic indexing while keeping all data and processing entirely on your local machine.74395MIT
- Alicense-qualityAmaintenanceProduction-ready MCP server providing RAG, hierarchical memory, and 8+ tools for AI agents via the Model Context Protocol.41Apache 2.0
- Alicense-qualityBmaintenanceAn MCP server that provides controlled read/write tools for managing local-first research memory in an Obsidian vault, enabling AI agents to maintain project context across sessions.71MIT
- AlicenseAqualityBmaintenanceMCP server for managing project memory in AGENTS.md. Allows agents to record and remove durable facts using save/forget tools.24541MIT
Related MCP Connectors
Persistent project context — Rust-native MCP server. IANA-registered .faf format.
Person-owned, portable AI memory as a remote MCP server, readable and writable by any MCP client.
Agent-native MCP server over the public saagarpatel.dev corpus. Read-only, stateless.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/Wolfe-Jam/faf-trinity'
If you have feedback or need assistance with the MCP directory API, please join our Discord server