APIClaw

The API layer for AI agents.

13,500+ installs. 26,704 discoverable APIs. 2,895 empirically callable (49 fully managed). Intelligent LLM Gateway.

APIClaw is in early release. Core features are live and actively expanding. Provider coverage, routing intelligence, and catalog depth grow with every update.

Install

curl -fsSL https://apiclaw.cloud/install.sh | bash

Adds APIClaw as an MCP server in your Claude, Cursor, or any MCP-compatible agent. Register your email to unlock 25 calls/month.

---

Four Ways to Use APIClaw

1. MCP Server (Agent Discovery + Calling)

Install APIClaw and your agent gets tools to discover and call APIs directly:

discover_apis("weather data for Stockholm")
-> Weatherstack, Visual Crossing, AccuWeather, OpenWeather...

call_api("frankfurter", "latest", {"from": "USD", "to": "SEK"})
-> { "rates": { "SEK": 10.85 } }

The agent handles everything through MCP tools. Works in Claude Desktop, Cursor, Windsurf, OpenClaw, and any MCP-compatible client.

2. Intelligent Gateway (OpenAI-compatible endpoint)

Generate an sk-claw- API key at apiclaw.cloud/workspace and use APIClaw as an LLM gateway from any application:

curl https://apiclaw.cloud/v1/chat/completions \
  -H "Authorization: Bearer sk-claw-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-4-6",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

3. Remote MCP — Grok, Cursor, ChatGPT, Claude (OAuth, zero install)

Paste one URL into your AI client. APIClaw handles the rest via OAuth.

https://apiclaw.cloud/mcp

Grok auto-discovers our authorization server, opens a sign-in window, you click "Authorize", and the connector is live. Standard MCP OAuth 2.1 (PKCE + RFC 7591 dynamic registration). No npm, no config file, no API keys to copy.

Connect to Grok in 20 seconds:

1. Open Grok → Settings → Connectors → Add MCP server
2. Paste https://apiclaw.cloud/mcp and click Connect
3. Sign in with your APIClaw email (or create the workspace) → Authorize

Done. Grok now has discover_apis, call_api, get_api_details, list_categories, list_connected, and check_balance.

Same flow works in Cursor, ChatGPT (custom GPT), Claude Desktop, and any MCP-compliant client — just point them at https://apiclaw.cloud/mcp.

Need pre-shared credentials instead of OAuth? Open Workspace → Integrations, click Generate connector, copy the Client ID and Secret.

Troubleshooting

Symptom	Fix
Grok doesn't auto-find APIClaw in connector search	Add it manually as a custom MCP server with URL https://apiclaw.cloud/mcp. We're rolling into the public catalogs in the next sweep.
Grok shows "Connected" but says it has no access when you ask it to use APIClaw	Start a fresh Grok chat — old chats keep their tool-call cache and don't pick up newly added connectors.
401 loop after authorize	Your access token expired or was revoked. Open Workspace → Integrations, revoke the connector, and re-add it in the client.
client_bound_to_other_workspace	A dynamic client is bound to the first workspace that approved consent. Sign in with the same email or use Generate connector to create a fresh one for the new workspace.
Consent screen redirects but client says "denied"	The browser blocked the redirect or you closed the tab early. Click "Authorize" again — auth codes are single-use but expire after 10 minutes so you can retry.

4. CLI

Humans in the terminal, scripts, CI. Ships with @nordsym/apiclaw:

npx @nordsym/apiclaw login                    # OTP auth, terminal-native
npx @nordsym/apiclaw setup                    # auto-detect Claude, Cursor, Windsurf
apiclaw discover "transcribe audio file"      # search registry
apiclaw-http                                  # stand up local HTTP gateway

All four interfaces route through the same apiclaw.cloud gateway. One billing pipeline, one logging pipeline, one registry. Email-verified workspace required for any token issuance — discovery is open, execution requires auth.

One endpoint. Automatic provider routing. The gateway selects the optimal provider based on your workspace settings:

Routing Mode	Behavior
balanced	Default. Weighs speed, cost, and quality
fastest	Lowest latency (Groq, Mistral)
best_price	Cheapest available provider
highest_quality	Best model quality
advisor	Opt-in. Mistral Small picks the optimal model per prompt (only fires when no model is set in the request)

Providers routed through the gateway:

• Groq -- Ultra-fast inference (Llama, Mixtral, Gemma)
• Mistral -- European models (Small, Large, Codestral)
• Together AI -- Open-source models (DeepSeek, Qwen, Llama)
• OpenRouter -- 800+ models as fallback (GPT, Claude, Gemini, etc.)

Override per-request with X-APIClaw-Route: fastest or target a provider directly: X-APIClaw-Route: groq.

The Intelligent Gateway is in its first release. Routing logic, provider coverage, and model support are actively expanding.

---

API Catalog

Browse all 26,704 indexed APIs at apiclaw.cloud/catalog.

• Search across 40+ categories
• Filter by tier: Managed (49), Callable (2,895), Discovery only (~23,800)
• Every "Callable" entry is empirically smoketested (HTTP 200 + parseable JSON in our last sweep)
• Infinite scroll, category filters, instant search

---

Discoverable vs Callable

Discoverable (26,704 APIs) -- Every API in the index. Your agent can search, read specs, and evaluate them. Free and unlimited.

Callable (2,895 APIs) -- Every entry empirically smoketested: returned HTTP 200 + parseable JSON in our last sweep. Two layers:

• Managed (49 directCallConfigs across 19 brands) -- APIClaw owns the keys. You call, APIClaw authenticates.
• Keyless verified (~2,846) -- Public APIs proxied through APIClaw, smoketest-passed. Sortable by verified latency.

The remaining ~6,500 OpenAPI-spec'd providers in the registry have valid integration paths but need context (POST body, path-vars, required query params) we couldn't smoketest blindly. They live in Discovery — your agent can still call them through APIClaw if it supplies the right inputs, but we don't claim them as callable until proven.

Managed Providers

Provider	What	Category
OpenRouter	800+ LLMs (GPT, Claude, Gemini, Llama)	AI & ML
Groq	Ultra-fast inference (Llama, Mixtral, Gemma)	AI & ML
Mistral	Mistral models (Small, Large, Codestral)	AI & ML
Together AI	Open-source models (DeepSeek, Qwen, Llama)	AI & ML
Cohere	RAG, reranking, embeddings	AI & ML
Replicate	ML models (Flux, SDXL, Whisper)	AI & ML
Stability AI	Image generation (SD3, SDXL)	AI & ML
ElevenLabs	Text-to-speech (29 languages)	Voice
Deepgram	Speech-to-text (Nova-3)	Voice
AssemblyAI	Audio intelligence, diarization	Voice
Brave Search	Privacy-first web search	Search
Serper	Google SERP results	Search
Firecrawl	Web scraping and crawling	Scraping
E2B	Code execution sandbox	Dev Tools
GitHub	Repository and code access	Dev Tools
APILayer	27 APIs (finance, geo, scraping, news)	Multi-API

---

MCP Tools

Tool	What
discover_apis	Search 26,704 APIs by capability
get_api_details	Full specs, pricing, auth
call_api	Execute through APIClaw proxy
list_connected	See available managed providers
check_balance	Usage and remaining calls
register_owner	Register email, unlock 25 calls/month

Pricing

Plan	Price	What you get
Free	$0 forever	Discovery (always free), 25 managed calls/month
Pay-as-you-go	Provider cost + 15% margin	All managed providers, no monthly fee, metered billing
Enterprise	Custom	Private deployment, custom limits, SLA, book a call

Pass-through pricing on managed provider calls + 15% margin. No hidden fees.

---

Catalog - Dashboard - Docs - Book a Call

MIT License