Repository navigation
Conversation
Every MCP session now registers the full TOOL_REGISTRY (33 tools). Removed the toolset-selection machinery that let operators or clients narrow the served surface: - T49_MCP_TOOLSETS / T49_MCP_READ_ONLY env vars - X-T49-MCP-Toolsets / X-T49-MCP-Read-Only request headers - packages/mcp/src/toolsets.ts (resolveToolSelection, TOOLSETS_HEADER) - toolsets/readOnly options on createTerminal49McpServer and ToolSession Header- or env-based hiding was not a real safety boundary: any client could send the headers itself. Write safety comes from the tier A/C confirmation protocol and tool annotations, which are unchanged. Toolsets remain documentation labels: whoami.enabled_toolsets reports populated toolsets, and TOOLS_OVERVIEW keeps the per-toolset grouping. Store/directory listings now match the served surface: - chatgpt-app-submission.json: +stop/resume shipment tracking, +cancel tracking request (annotations + justifications) - claude-connector-submission.json: tool_names +3 Docs updated to the all-tools surface (README table merged to 33 rows, TOOLS_OVERVIEW, mcp.mdx, home.mdx, MCP_README, WORKOS_MCP_SETUP, .env.local.example). Snapshots updated intentionally.
Contributor
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Automations to automatically generate PRs for you. |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
Shadow risk calibration
Calibration result: agreement. Shadow risk never blocks or enables merge. Keep it in observation mode until roughly 50 pull requests have human-reviewed labels, then calibrate before considering any policy use. |
|
Found 2 test failures on Blacksmith runners: Failures
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
![Fix with [code]smith](https://pr-comments-assets.blacksmith.sh/codesmith/fix-with-codesmith-light-v2-3x.png)
What
Every MCP session — shared token or signed-in user, any client — now gets the full registered tool surface (33 tools). This removes the machinery that let an operator env var or a request header narrow what a session sees:
T49_MCP_TOOLSETS/T49_MCP_READ_ONLYenv varsX-T49-MCP-Toolsets/X-T49-MCP-Read-Onlyrequest headerspackages/mcp/src/toolsets.ts(resolveToolSelection, header parsing) and its testtoolsets/readOnlyoptions oncreateTerminal49McpServer()andToolSessionWhy
Header- or env-based tool hiding was not a real safety boundary: the headers are sent by the client, so any client could send them itself and get whatever surface it asked for. It also made the tool surface differ per deployment/client, which made docs, store listings, and evals drift from what users actually saw.
Write safety is unchanged and comes from the right layer: the tier A/C confirmation protocol (preview → approve → confirm), tool annotations (
destructiveHint,readOnlyHint), andT49_MCP_CONFIRMATION_SECRETfor unattended HTTP writes. Destructive tools still require a confirmed preview; nothing about authorization or the confirmation flow changes.Toolsets remain as documentation labels:
whoamistill reportsenabled_toolsets(now the populated toolsets), andTOOLS_OVERVIEW.mdkeeps its per-toolset grouping.Store / directory listings (now consistent with the served surface)
chatgpt-app-submission.json: addsstop_shipment_tracking,resume_shipment_tracking,cancel_tracking_requestwith annotations + justifications (the parity test enforces this).claude-connector-submission.json:listing.tool_names+3.The ChatGPT and Claude directory listings are already live, so after this merges the portal listings need a refresh to add the three tracking_lifecycle tools and match what the server serves.
Tests
tsc --noEmitclean.instructions.txt,tools-list.json) updated intentionally: instructions and tools/list now advertise all 33 tools.Part of the MCP toolset-hiding cleanup; write safety stays in the confirmation protocol, not in hidden tools.