Tools and MCP
Tools let an agent read information or perform actions. MCP (Model Context Protocol) lets Maestrly connect to tool and resource servers on your device or at a network endpoint.
Add an MCP server
Open the tools or MCP settings and add the connection type provided by the server:
- for a local server, configure its executable, arguments, and required environment values;
- for a remote server, configure its URL and complete the supported authentication flow.
Start the connection and inspect the tools and resources it exposes. If startup fails, run the server's documented health check and verify its command, URL, credentials, and network access.
Review access before enabling
For every tool, ask:
- What data can it read?
- What can it create, change, send, or delete?
- Does it call a remote service?
- Which account and permissions does it use?
- Can consequential actions require your approval?
Enable the smallest set needed. A read-only task rarely needs write or administrative access.
Use tools in a chat
Attach the server or allowed tools to the agent or chat. State any important scope in the request—for example, a repository, database, workspace, or date range. Watch the first calls and confirm that arguments and targets are correct before allowing broader execution.
The selected AI model must support the relevant tool-calling behavior. The server can also be unavailable, rate-limited, or restricted by its own account plan.
Understand the data path
- Local servers run on your machine but may still contact external services.
- Remote servers receive the request data needed to execute their tools.
- The chosen AI provider receives the prompt, tool descriptions, relevant context, and tool results needed for inference.
- Maestrly has no hosted backend or inference/MCP proxy. Enabled integrations connect to their selected destinations.
Review each provider's and server's privacy terms before sending sensitive data.
Troubleshoot safely
- Server does not start: verify the executable exists, arguments are valid, and required environment values are present.
- Authentication fails: reconnect using the server's own account and narrow permissions where possible.
- Tool is missing: refresh the connection and verify the server version and model support.
- Call times out: check network access, server health, and provider rate limits before retrying.
- Unexpected action: stop the run, revoke or disable the tool, inspect logs, and rotate exposed credentials if needed.
Next: Agents and skills or Projects and Git.