Skip to main content
Tools let an agent act: read a repository, send an email, query a database. In Tilde, the gateway executes tool calls on the agent’s behalf. It injects the credential from a connection, checks the agent’s capabilities, and records the call. The agent never holds a provider credential. It receives tool definitions, and it calls them with its short-lived invocation token.

Types of tools

Tilde supports four sources of tools. You can mix them on one agent.

Managed tools

Managed tools are off-the-shelf integrations for providers such as GitHub, Google Workspace, Slack, Sentry, Stripe, and AWS. You connect an account once, and Tilde keeps the credential encrypted. Tilde injects it only when it invokes the selected tool. In Tools, select Configure Tools, choose a provider, and connect an account. Authentication is selected for each configured account, and OAuth tokens rotate automatically.

Proxied MCP servers

A proxied MCP server connects an existing Streamable HTTP MCP server to Tilde. The upstream server runs the tools. Tilde holds the OAuth, API key, or bearer credential, and stands between the agent and the server. In Tools, select Proxied MCP servers, then Browse provider catalog, or enter your own server’s URL. Tilde shows you the server’s tools before you connect credentials, and keeps them in sync with the server afterwards. Use this for internal MCP servers too. Your agents then reach them through one audited path, rather than each holding a credential.

Remote custom tools

Use a remote custom tool when the implementation should run independently of the agent. Your server exposes a discovery manifest and an invocation endpoint. Tilde signs both requests with the custom tool provider’s signing key, so your server can verify that a call came from Tilde. In Tools, select Custom Tools, then Remote HTTP server, and enter your endpoint. Tilde shows the signing key once. Store it as a secret on your tool server.

Local custom tools

Use a local custom tool when it needs in-process state, request context, or code that should not be deployed separately. You define it in your agent with your framework’s own tool type, and pass it alongside the tools from Tilde.
Local tools execute in the agent process. Tilde does not proxy them, so their credentials and other sensitive inputs remain your responsibility. Prefer managed, proxied, or remote tools for anything that touches a secret.

Group tools in MCP servers

You group the tools an agent can use into an MCP server in Tilde, and then connect the agent to it. One MCP server can combine managed tools, proxied MCP servers, and remote custom tools. We recommend dynamic mode for servers with many tools. In dynamic mode, the agent searches for tools and loads their schemas on demand, which keeps its context small.

Use another agent as a tool

Any registered agent can be a tool for another agent. Tilde exposes it as a tool that sends a message and waits for the response. Grant the calling agent the agents.invoke capability for the target agent.

Control tool access

Tool access is deny by default.
  • Capabilities. An agent can invoke only the tools named in its tools.invoke capability. Set it to none, all, or a selected list in the agent’s Capabilities tab. See IAM.
  • Visibility. Tool accounts and MCP servers can belong to a workspace or to one person. Visibility of an MCP server does not authorize a private tool group inside it. Tilde intersects both at execution time.
  • Human approval. Add the Tilde Human Approval tools to an MCP server when an action needs sign-off. The agent requests approval and waits until a person approves, denies, or lets the request expire.
  • Security middleware. Text an agent sends through a channel tool passes through response guards first.

Audit tool calls

Tilde records every proxied tool call against the invocation that made it. Review calls in the agent’s Sessions tab, and their timing in Traces. The audit trail records tool input after security middleware has screened it.