Back to directory

FastAgent

A serving runtime that takes an existing file-defined agent directory and runs it as a live service, embedded in a web app, deployed as a Telegram, Slack or Feishu bot, or exposed as an API endpoint.

FastAgent

FastAgent does not introduce a new agent-authoring format; a developer brings an existing agent directory, with its markdown instructions, reusable skills and TypeScript tools kept as ordinary files, and FastAgent provides the serving runtime and channel adapters around it. It can serve the same agent as a Telegram, Slack, Feishu or Lark bot, a verified webhook handler, or a streaming HTTP/SSE endpoint, or be mounted directly as one route inside a Next.js, Astro, Hono, Bun or Node application using the developer's own authentication, database and infrastructure.

Any model provider, including OpenAI, Anthropic or Google, can be used through OAuth or an API key, with typed tools discovered automatically from a tools folder and validated with Zod, and Agent Skills loaded on demand. Deployment targets, generated by a single command, include local Docker, Fly, Railway or AWS Bedrock AgentCore, each with a matching runbook; a local Docker deployment can also add a temporary tunnel for webhook testing.

It is built on the open-source pi agent harness, is written in TypeScript and released under the MIT license.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.