agentbull/ankole/lark-im

lark-im

Send, read, update, and withdraw Feishu or Lark messages and cards; manage reactions, chats, and attachments.

Find the operation in the operation index, then read only its linked JSON contract for inputs, outputs, and retry annotations.

Write the JSON argument object to a temporary file, then pass it over stdin so quotes and shell metacharacters remain data:

mcporter call 'lark_im.<tool-name>' --json - --output json --timeout 360000 < /absolute/path/arguments.json

After the call finishes, delete the temporary argument file and check the process exit code before parsing stdout.

Run mcporter list to discover the available connections. The default server is lark_im. Select lark_im@<connection-slug> for a particular application, or copy the exact lark_im@<connection-slug>@<account-key> server name from the list to select one application account. Account keys in server names are URL-encoded. If the intended workspace is unclear, ask the user to choose; do not switch accounts to work around an authorization failure.

Applet routes and Actions use await AnkoleRuntime.mcp.call(tool, args, { server: "lark_im", connection, account }) with the same contract. Omit connection/account only when the intended default is unambiguous. AnkoleRuntime.admin.mcp.call represents the current page owner; it does not select an application account.

Use stable user_id for people across the tenant's applications. Supply principal_uid when addressing a known Ankole colleague. Use file_ref in the form pages/<page-id>/<path> for files; attachments stream through the control plane. Message content accepts a structured object. im_send_message and im_reply_message have optional uuid; reuse it if the caller already has a stable request ID.

If no suitable connection or application account is listed, direct the user to the tenant’s Connectors page; a tenant administrator connects an application workspace, and members connect their own personal accounts.

For a file created in the Worker workspace, use the Pod files.upload operation with the target page ID, local_path, and role before calling an upload tool. Follow the pod-content Skill to choose or create that page and handle file versions. Use the returned page_id and file.path as pages/<page-id>/<path> in file_ref; a Worker filesystem path is not a Pod file reference.

An oauth_consent_required result identifies a missing personal authorization. Present the returned consent link, then repeat the intended operation after authorization. upstream_outcome_unknown means a write may have completed; inspect its result before repeating it. Respect retryAfterSeconds and use continueSlice for long waits in Applet Actions. In Applet error handlers, read recovery metadata such as authorization_url, retryAfterSeconds, and a partially created resource_id from error.details; error.code identifies the failure.