FastClaw
FastClaw
A lightweight AI Agent OS written in Go: Putting Agent Factory, model providers, system files, long-term memory, sandbox tools, Skills, Channels, plugins, upstream API and project preview runtime into the same deployable backend.
Single binary
Start the Dashboard and Agent backend with one local command
/v1 API
OpenAI-compatible connects to your own SaaS
Agent OS
Unified scheduling of tools, memories, channels and runtimes
Workflow
FastClaw
Operate from Dashboard
Manage Agents, Skills, Models, API Keys, Channels, Scheduler and Settings from the local console.
Expose capabilities to products
The upstream application only calls the /v1 API, FastClaw takes care of sessions, tools, memory and execution.
Use channels to expand reach
The same Agent can enter Web chat, Slack, Telegram, Discord, Feishu, LINE and other channels.
Extend the platform with plugins
JSON-RPC plugins can contribute tools, providers, hooks, and messaging capabilities.
Prompt pattern
Write SOUL.md, IDENTITY.md and USER.md templates for this customer service agent, limiting it to only answer product documents and order status.
How to use it
Understand the tool, then read the docs
This entry page keeps the tool overview, use cases, and first steps up front, then organizes the documentation index for deeper reading.
Agent Factory
Each Agent has SOUL, IDENTITY, USER, MEMORY, skills, tools, sessions and model configurations.
Upstream API
Use /v1/chat/completions, app_user, usage and quota to connect Agent capabilities to the product backend.
Tools and Sandbox
It has built-in tools such as files, commands, networks, memories, and project previews, and uses policy to limit execution boundaries.
Project Runtime
Launch template applications for coding agents, preview pages, read logs, and let HMR reflect real-time changes.
Install single binary
First start FastClaw on the local machine and complete the first setup and provider configuration.
Create Agent
Use Dashboard or CLI to create Agent, select provider/model, and write SOUL and system files.
Connect the product API
The server calls /v1/chat/completions with API key and passes in end user and session key.
Tighten governance boundaries
Check API key, Agent ACL, sandbox policy, channel webhook, quota, logs and backup before going online.
FastClaw Help Docs
Documentation index
Chinese help documents compiled based on FastClaw open source source code: Agent OS architecture, Dashboard, configuration, tool sandbox, Skills, Channels, plug-ins, upstream API, project runtime, deployment and governance.
Get started
FastClaw Overview
FastClaw is a lightweight AI Agent OS written in Go that puts Agent Factory, sessions, tools, sandboxes, skills, channels and upstream APIs into the same runtime.
ReadGet started
quick start
From installing a single binary to entering the Dashboard: run it first, then create the provider, agent and first session.
ReadGet started
console
FastClaw Dashboard is a local console: manage Agents, Skills, Models, API Keys, Settings, and access each Agent’s chat, files, channels, and scheduling panels.
ReadGet started
Overview
FastClaw is a lightweight AI Agent OS written in Go that puts Agent Factory, sessions, tools, sandboxes, skills, channels and upstream APIs into the same runtime.
quick start
From installing a single binary to entering the Dashboard: run it first, then create the provider, agent and first session.
console
FastClaw Dashboard is a local console: manage Agents, Skills, Models, API Keys, Settings, and access each Agent’s chat, files, channels, and scheduling panels.
daemon
Understand the differences between fastclaw foreground operation, daemon background operation, service installation and local ports.
Run the model
Architecture
The core architecture of FastClaw is Gateway + Store + Agent Manager + Tool Registry + Channel Manager + Project Runtime.
Configuration
FastClaw does not have fastclaw.json: startup guidance relies on FASTCLAW_* environment variables, long-term configuration into the database, and can be modified through Dashboard or CLI.
agent
Agents are the core objects of FastClaw: models, system files, memories, skills, tools, sessions, sandboxes, and channel bindings are all organized around Agents.
model supplier
FastClaw supports OpenAI, Anthropic, Ollama, OpenRouter, Groq, DeepSeek, Mistral, and OpenAI-compatible APIs.
Ability level
Tools and Sandbox
FastClaw has built-in tools such as files, commands, networks, memories, messages, subagents, Cron, preferences, project previews, etc., and uses sandbox policy to control execution boundaries.
Skills and Memory
Skills is a reusable capability package, and Memory is a long-term context layer; together, the two allow the Agent to not only chat, but also to stably execute the process.
channel
Channels bind the same Agent to web chat and team messaging portals, and isolate context by channel session.
scheduler
FastClaw has a built-in Cron scheduler that allows the Agent to trigger tasks based on time and create, list, and delete scheduled tasks in sessions.
plug-in
The FastClaw plug-in is a JSON-RPC subprocess that can register tools, providers, hooks, and channel capabilities, and declare capabilities through manifests.
APIs and integrations
Upstream API
FastClaw provides /v1/* API for upstream applications. The core is OpenAI-compatible chat completions, plus agents, users, usage, and quota.
User and Key
The key to production access to FastClaw is not to just issue an API key, but to design app_user, end user, Agent ACL, quota and session isolation.
project runtime
Project Runtime is FastClaw's coding agent layer: projects, running instances, template scaffolding, dev server, previews, logs, and changed files.
Operations, Maintenance and Governance
deploy
FastClaw can be run locally on a single machine, or can be used for production deployment using Postgres, Redis, object storage, and Docker/E2B/Boxlite sandboxes.
command line
The fastclaw CLI is a local control portal: agent, apikey, provider, plugin, policy, projects, sandbox, session, skill, tools, usage, and version all have subcommands.
safe
FastClaw's security boundary consists of API key, Agent ACL, sandbox policy, secret scrub, channel verification, quota and log audit.
license
FastClaw uses the FastClaw Community License: based on Apache 2.0, with the addition of multi-tenant FastClaw-as-a-Service, branding and trademarking related restrictions.