Skip to content

Separate Roo Code tool discovery from Montage execution permission

Roo Code can display tools before an authorized workspace call succeeds.

Integrations

How do I separate Roo Code tool discovery from Montage execution permission?

Roo Code can display tools before an authorized workspace call succeeds. Use a status-only review that treats the discovered interface as metadata and refuses to widen automatic execution just to make a failed request appear successful or complete. The tool list is a contract description, not a workspace receipt. Keep automatic execution narrow and require an actual authorized status response before making any claim that this Roo session obtained private account access.

01 / CONFIGUREOne remote endpointUse this client's documented HTTP setup
02 / AUTHORIZEYour workspace accessUse a supported sign-in or credential flow
03 / VERIFYRun signals_statusCheck a real workspace response
Copyable task-specific request
Help me separate Roo Code tool discovery from Montage execution permission. Ask whether this Roo session can read the intended workspace, not merely display tool names. Report listed, authorized and executed as separate states. No writes.
01 / Integrations

Name the diagnostic question

Ask whether this Roo session can read the intended workspace, not merely display tool names. Leave research and mutations outside the investigation’s permitted effects. A real readiness question names the expected account-scoped fact; it does not ask the agent to demonstrate every advertised server operation.

02 / Integrations

Confirm the effective target

Inspect project overrides and the active remote endpoint. A correct global definition cannot establish the selected target when .roo/mcp.json sends this project elsewhere. The active entry should be traced to its owning configuration so a project override cannot be overlooked when selecting the status tool.

03 / Integrations

Review authorization eligibility

Proceed only with compatible OAuth or an existing approved owner-controlled path. Discovery does not authorize attempts to access private operations through an unsupported authentication arrangement. Unsupported consent is not repaired by accepting broader automatic approvals, because user approval and server account authorization answer different security questions.

04 / Integrations

Select the read explicitly

Request signals_status rather than letting the agent test arbitrary tools. Keep approvals narrow and do not enable broad automation to eliminate a transport or authorization warning. Do not let a failed read become permission to execute another operation chosen by the agent merely because it looks easier to test.

05 / Integrations

Classify evidence accurately

Report listed, authorized and executed as separate states. Without a real status result, leave workspace access unverified and use the approved fallback instead of certifying Roo connectivity. The evidence summary should distinguish tool discovery from real authorized execution and clearly identify any approved alternative used for the user’s underlying need.

Keep the boundaries clear.

  • Unsupported consent is not repaired by accepting broader automatic approvals, because user approval and server account authorization answer different security questions.
  • The evidence summary should distinguish tool discovery from real authorized execution and clearly identify any approved alternative used for the user’s underlying need.

Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.

Primary references

Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.

Know what social media shows about you.
Start with a question.

Get started
montage