Skip to main content

ClawPod Guide tools for Agents

clawpod-guide is a read-only MCP server registered in ClawPod Agent images. It remotely reads main of the public Wondermove-Inc/clawpod-guide repository without cloning it or saving documents in the Agent workspace.

It requires a running image containing the server and runtime network policy allowing GitHub. Existing Agents do not gain it until replaced/restarted with a supporting image release.

Registration​

The image's default MCP configuration registers clawpod-guide automatically. No Portal API key or manual MCP server address is needed. First check whether it appears in the current tool list.

If it is missing or GitHub is unreachable, do not search for installation files or copy documentation into the workspace. Report the available tools and failure and narrow the task to user-provided documents or other permitted sources.

APIs​

search_clawpod_guide​

Find relevant pages. Each call reads the public repository's search index and returns paths, titles, descriptions, and relevant context.

InputRequiredMeaning
queryYesNatural-language feature, setting, or task
limitNoDefault 5, maximum 10 results

Search results alone do not establish setup, permissions, or support. Read the returned page next.

read_clawpod_guide​

Read lines from a page returned by search or section listing. Arbitrary URLs and paths outside the repository are not accepted.

InputRequiredMeaning
pathYesExact returned path, such as guides/manage-agent
start_lineNoFirst line to read; default 1
max_linesNoDefault 200, maximum 400

If the response includes next_start_line, use it with the same path to continue.

list_clawpod_guide_sections​

List sections and pages. Supply group to narrow to one section. Use this when the query or documentation scope is unclear.

Agent workflow​

  1. Search for relevant pages.
  2. Read the best match for procedure, prerequisites, and limits.
  3. Prioritize current authorization, organization policy, and actual tools/permissions over documentation when assessing execution scope.
  4. If a page or runtime tool is missing, do not conclude the capability is impossible; follow safe interpretation.

The tools only read documentation. They do not access or modify secrets, tenant data, Agent settings, or external services.