FastAgent CLI
FastAgent turns an agent directory with skills and TypeScript tools into a live service with bots and schedules
- 66 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent answers in Telegram, Slack and via webhooks, so channel tokens and model keys sit on the server
- TypeScript tools run code on the server and can change application data
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/fastagentDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent answers in Telegram, Slack and via webhooks, so channel tokens and model keys sit on the server
- TypeScript tools run code on the server and can change application data
This is third-party code. Review the repository files before installing.
What it does
FastAgent is a TypeScript CLI and library that takes an agent out of the terminal and runs it as a live service. The agent stays a directory: a `SYSTEM.md` prompt, skills and typed tools in `src/tools/` are files you can edit and commit. There is no new DSL; it consumes standards such as `AGENTS.md` and Agent Skills. Channels connect the agent to Telegram, Slack and Feishu, to a schedule or to your own adapter: webhooks are verified, replies stream and groups are handled. Conversations live in the instance storage and an accepted message is answered even after a restart. You can open the agent inside your own Next, Astro, Hono or Node app and call it like a function behind your auth. The `fastagent init`, `fastagent dev` and `fastagent invoke "hello"` commands create the agent, serve it and send the first message. `fastagent deploy docker|fly|railway|agentcore` generates a container, target config and a runbook, and `--run` drives the deployment to the end. It works with any supported model provider through OAuth or an API key.
Who it is for. Developers who built an agent in the terminal with a coding agent and want to take it out into a bot, a webhook or a call from their own app.
Good fit when
- An agent from a directory has to answer in Telegram or Slack and keep conversations across restarts
- The agent has to sit inside a Next or Hono app and be called behind your auth and database
- Once a day the agent has to start on a schedule and post a summary
- You need a one-command deploy to Docker, Fly, Railway or AWS Bedrock AgentCore
Not a fit when
- You need a workflow engine with fixed steps: the agent chooses its own order, and orchestration belongs in your own queue
- You need a dashboard and user roles: FastAgent has none and has one key per instance
- The server has no Node.js 24.15 or Bun 1.4
Example request
Create a support agent with fastagent init, start fastagent dev and send it a message with fastagent invokeLimitations
- It needs Node.js 24.15 or newer or Bun 1.4 or newer, because the agent storage uses `node:sqlite`.
- The project is pre-1.0 and its author expects rapid changes and experimental features.
- `FASTAGENT_API_KEY` is one key per instance and stays on the server; users, roles and per-session access belong to your app.
- It builds on the pi harness and brings no dashboard or control layer: a stable ingress for webhooks is yours to provide.
How to disable. Remove the package with `npm uninstall -g @fastagent-sh/fastagent`; the agent directory stays as ordinary project files.
FAQ
How do I install FastAgent?
Run `npm i -g @fastagent-sh/fastagent` for the CLI and `npm i @fastagent-sh/fastagent` for the library. It needs Node.js 24.15 or Bun 1.4.
How do I get a first agent running quickly?
Run `fastagent init my-agent`, enter the folder, start `fastagent dev` and send `fastagent invoke "hello"`. `fastagent start` runs production-style locally.
Where can I deploy an agent?
`fastagent deploy` supports Docker, Fly, Railway and AWS Bedrock AgentCore. It generates config and a runbook for each target, and `--run` finishes the launch.
Do I have to rewrite my agent for FastAgent?
No. It reads the existing directory: `SYSTEM.md`, skills and TypeScript tools, and adds a serving layer around them.
Which models are supported?
Any provider that pi-ai supports, including OpenAI, Anthropic and Google, via OAuth or API key; you can write your own provider in code.
Discussion
No questions or reviews yet.
Related
CLI toolsAn open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
A scaffold of personas, rules and skills in plain Markdown: Maestro hands work to different models and spares your main plan
A CLI that keeps instructions, commands, skills and MCP in .agents and links them into the configs of different agents
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
The squads CLI runs teams of agents defined in markdown files in git: a lead briefs, workers act, a verifier checks.
README in short
The README explains the gap between a terminal agent and a live service, lists features and what FastAgent leaves out, and gives the install, a quickstart and an embedding example.