Build with Agentik
Give your application or agent one connection to tools, with visible prices and a shared prepaid balance.
Start with a read
Explore the tool catalog, then inspect an exact tool before executing it. Discovery and inspection do not start a provider execution. A workspace and an authorised credential are required to run a tool, including a tool priced at zero.
This environment publishes 3,539 tools from 165 providers.
Choose your connection
- MCP and REST quickstart: discover, inspect, then authenticate.
- Connection guides: Claude Desktop Chat/Cowork, Claude Code Desktop or terminal, ChatGPT, Codex, Cursor, Hermes and custom API/CLI runtimes such as Muse.
- Public OpenAPI contract: inputs, results and errors.
- Installation instructions: configuration tailored to this environment.
- Agent skill: the discover, inspect and run loop, prices, caps and webhooks in one SKILL.md for Claude Code and other agents.
- Claude Code plugin:
/plugin marketplace add agent-tool-router/agentik, then/plugin install agentik@agentik. Source (MIT): github.com/agent-tool-router/agentik.
Use Agentik from your terminal or your code
Run tools from a shell or from your own application, on the same catalog, prices and prepaid balance.
CLI
Connect a client, install the agent skill, and discover, inspect and run tools from a shell.
TypeScript SDK
Discover, inspect and run tools from your code, with adapters for OpenAI, Anthropic and the AI SDK.
Both are open source (MIT): github.com/agent-tool-router/agentik
Keep execution under control
Read authentication, prices and reservations, idempotency, asynchronous results and error recovery. Set optional spending and tool restrictions for each key in the workspace.
For agents
Start at llms.txt. Every public document and catalog page has a Markdown link; request Accept: text/markdown or append .md. Capabilities and pricing are generated from the same backend contract.
Updated
Connect your agent to 3,500+ tools for free
Sign in with Google or GitHub. No card required.