Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

FastClaw is a lightweight AI Agent runtime written in Go, positioned as an “Agent Factory” — it creates, manages, and runs multiple AI agents, each with its own personality (SOUL.md), memory, skills, and tools. FastClaw handles LLM communication, tool execution, sandbox isolation, and session management out of the box, ships as a single binary, and comes with a built-in Web Dashboard. By integrating APIYI, you get:

🚀 Single-Binary Deployment

One-line install, bundled SQLite, runs locally or in the cloud

🤖 Multi-Agent Management

Each agent has its own personality, model, skills, and sessions

📱 Multi-Channel IM

Built-in Telegram / Discord / Slack channel bindings

🛡️ Sandbox Isolation

Docker / E2B sandbox support for secure tool execution
Project Info: FastClaw is source-available under the FastClaw Community License (Apache 2.0 + addendum). Repository: github.com/fastclaw-ai/fastclaw.

​
Install & Start

Install FastClaw with the official one-liner — it drops a single binary into ~/.local/bin: 
curl -fsSL https://raw.githubusercontent.com/fastclaw-ai/fastclaw/main/install.sh | bash
The first run launches a setup wizard. After configuring your LLM provider, a default agent is created automatically: 
fastclaw                    # Foreground (Ctrl+C to stop)
fastclaw daemon start       # Background (logs at ~/.fastclaw/daemon.log)
fastclaw daemon install     # Register as a launchd / systemd service
Then open the dashboard at http://localhost:18953 (default port 18953). APIYI is compatible with both the OpenAI and Anthropic API protocols. FastClaw can use either an OpenAI-compatible Provider or an Anthropic-compatible Provider — both reach the same full model matrix with a single APIYI key.
  1. Open http://localhost:18953 and log in with the admin account generated on first run
  2. Go to Models / Providers and add a new Provider entry:
FieldRecommended Value
Provider typeOpenAI compatible
Base URLhttps://api.apiyi.com/v1
API KeyYour APIYI key (sk-...)
ModelsAdd as needed, e.g. gpt-5.4, claude-sonnet-4-6, deepseek-v3.2, gemini-3.1-pro-preview
  1. Go to the agent’s Models panel and set it as the default model

​
Option 2: Configure via CLI

# 1. Create a new agent, initially bound to APIYI (OpenAI-compatible)
fastclaw agents init alpha \
  --provider openai \
  --model openai/gpt-5.4 \
  --api-key-env APIYI_API_KEY

# 2. Point the OpenAI provider's Base URL at APIYI
fastclaw agents config alpha set provider.openai.apiBase https://api.apiyi.com/v1
fastclaw agents config alpha set provider.openai.apiKeyEnv APIYI_API_KEY

# 3. Add the models you want (append, idempotent)
fastclaw agents config alpha set provider.openai.model gpt-5.4
fastclaw agents config alpha set provider.openai.model claude-sonnet-4-6
fastclaw agents config alpha set provider.openai.model deepseek-v3.2
The APIYI_API_KEY environment variable must be exported in your shell first (export APIYI_API_KEY=sk-...) — FastClaw does not store the key in plaintext in the database.
Why APIYI?
  • One key, many models: OpenAI / Anthropic / Google / DeepSeek / Zhipu and more — no need to apply for each provider separately
  • Pricing advantage: typically 5%-20% off vs. official pricing, with deposit bonuses on selected models
  • Direct access in China: reach overseas LLMs without a VPN
  • Dual-protocol compatibility: both FastClaw provider types are supported

​
Option 3: Anthropic Native Protocol (Best for Claude-Heavy Workloads)

If you primarily use Claude models, point the Anthropic Provider directly at APIYI:
FieldRecommended Value
Provider typeAnthropic
Base URLhttps://api.apiyi.com
API KeyYour APIYI key
Modelsclaude-sonnet-4-6, claude-opus-4-7, etc.
Equivalent CLI: 
fastclaw agents config alpha set provider.anthropic.apiBase https://api.apiyi.com
fastclaw agents config alpha set provider.anthropic.apiKeyEnv APIYI_API_KEY
fastclaw agents config alpha set provider.anthropic.model claude-sonnet-4-6
fastclaw agents config alpha set model claude-sonnet-4-6

​
Feature Cheat Sheet

Agent Management

Dashboard → Agents: create / edit agents, define SOUL.md (personality), IDENTITY.md, MEMORY.md (long-term memory)

Skills

Bundled: code-runner, image-gen, data-analysis, web-search, skill-creator. Install more from ClawHub / GitHub.

IM Channel Bindings

Agent → Channels: paste Telegram / Discord / Slack bot tokens — auto-validated on save.

OpenAI-Compatible API

/v1/chat/completions streaming endpoint works with any OpenAI SDK out of the box.

Sandbox Execution

Settings → Runtime to toggle Docker / E2B sandbox, with auto-sync of artifacts after tool calls.

Scheduler

Agent → Scheduler: let the agent create cron-based reminders via create_cron_job.

​
Deployment Modes

ModeUse CaseKey Config
LocalPersonal usefastclaw daemon start, default SQLite storage
DockerSingle-host servicecd deploy/docker && ./start.sh
KubernetesMulti-replica prodFASTCLAW_STORAGE_TYPE=postgres + S3 object store
Multi-replica deployments require:
  • FASTCLAW_STORAGE_TYPE=postgres, FASTCLAW_STORAGE_DSN=postgres://...
  • A set of FASTCLAW_OBJECT_STORE_* variables pointing at an S3-compatible store (used to sync skills and workspaces across pods)
  • FASTCLAW_BIND=all (listen on 0.0.0.0)
Full K8s manifests live in the repo’s deploy/k8s/ folder.

​
FAQ

  • FastClaw: Go-based, single binary, geared toward the “Agent Factory” use case — multi-agent management, IM channel delivery, sandbox isolation. Best when you want to ship agents as a service.
  • OpenClaw: Node.js-based, focused on the personal local-assistant use case — local privacy + multi-IM-platform interplay.
  • Both can reach APIYI’s full model matrix. Pick whichever fits your deployment shape.
Yes. APIYI exposes both an OpenAI-compatible endpoint (https://api.apiyi.com/v1) and an Anthropic-compatible endpoint (https://api.apiyi.com). Either FastClaw provider type works — use the model IDs from APIYI’s docs directly (e.g. gpt-5.4, claude-sonnet-4-6, deepseek-v3.2, gemini-3.1-pro-preview).
  1. Create a bot on the target platform and obtain a token
  2. Dashboard → pick the agent → Channels → paste the token and save (it’s auto-validated via getMe / auth.test)
  3. Search the bot inside the IM platform and start chatting — sessions are isolated per chatID
Yes. The FastClaw Community License allows:
  • ✅ Embedding as the backend of your own product (commercial use)
  • ✅ Internal deployment within your organization
Not allowed (without a commercial license):
  • ❌ Hosting FastClaw itself as a multi-tenant SaaS to unrelated organizations
  • ❌ Removing or altering FastClaw branding in the dashboard
Commercial licensing: [email protected]

Project Repository

github.com/fastclaw-ai/fastclaw

APIYI API Docs

Get your key and Base URL reference

OpenClaw Alternative

A different option for the personal local-assistant use case

Pricing & Top-Ups

See APIYI pricing and first-recharge bonuses