Claudekit — Claude Code with a senior developer watching over it
Claude Code ships dangerously by default. Claudekit gives it built-in protection, project memory, and specialist subagents — so it can't make big mistakes silently, doesn't forget what you decided yesterday, and says done when it actually is.
https://claudecode-kit.com/Opens ChatGPT on the web or desktop and asks it to use the WebMCP tools available here.
Connect straight to this server’s public endpoint.
https://mcp.claudecode-kit.com/mcpWe add this server to your workspace, then open Studio — saved access, one connection to many servers, with a history of what ran.
Last probed Sep 14, 2026 · mcp.claudecode-kit.com
27tools discovered
Showing 25 of 27 from the live probe.
Whoami
Return identity info for the calling license. ADR 0011 § 6 shape: authenticated, license_id_last6, hash_last6, plan, status, is_active, project_id (None until T007.1), tier (Free/Pro/Unauth), trial_state (not-yet-implemented until migration 0015).
Record Plugin Event
Record an observable plugin event for the calling license. Used by the claudekit plugin to surface activity on the personal-area dashboard. Payload must not contain credentials, tokens, or secrets — the validator rejects forbidden keys. Returns the event_id and recorded_at timestamp.
Log Telemetry
Push an arbitrary observability line to the dashboard's telemetry feed for the calling license. Use this for ad-hoc debug/info/warn/error events whose shape doesn't justify a dedicated event_type — e.g. "audit.started", "doctor.check_passed". Lands as event_type='telemetry' in plugin_events; the dashboard renders it in the live feed with the supplied level and payload. Same forbidden-key validator as record_plugin_event.
Start Onboarding
Begin claudekit onboarding for a project. Returns the first question and the tool to call next.
Reload Plugin
Signal that the claudekit plugin has been freshly loaded. Call immediately after start_onboarding returns state='plugin_installed'. Returns state='plugin_reloaded' + next_tool='check_entitlement'.
Check Entitlement
Resolve the caller's license tier (free / monthly / yearly) and record it on the session. Call after reload_plugin. Returns state='entitlement_checked' + entitlement_tier + next_tool='route_entitlement'.
Route Entitlement
Branch the onboarding flow based on the caller's license tier. Free and monthly/yearly plans both proceed to advance_to_onboarding. (Pro activation prompt is T012 scope — not yet implemented.) Returns state='free_or_activation_prompt' + next_tool='advance_to_onboarding'.
Advance To Onboarding
Advance the session from free_or_activation_prompt to onboarding_questions; the kit is ready to begin asking installation questions.
Record Project Type
Record whether this is a new, existing, or mixed project. Call after start_onboarding.
Record Known Stack
Record the project's tech stack when the user knows it. Call after record_project_type, instead of detect_stack.
Detect Stack
Run automatic stack detection on the project files. Call after record_project_type when the user says 'detect'.
Advance Onboarding Detection
Advance the onboarding state after stack detection has run. Pass the DetectionResult returned by detect_stack.
Record Kit Preset
Record whether the user wants the mini or full kit preset. Call after stack is known.
Record Hooks Optin
Record whether the user opts into Claude Code hooks. Call after record_kit_preset.
Cancel Onboarding
Cancel an in-progress onboarding session. Call this if the user wants to stop at any point. State advances to 'cancelled' (terminal); no kit files are written. Call start_onboarding again any time to begin a fresh session.
Advance To Manifest Preview
Advance the session from onboarding_questions to manifest_preview. Call this only when all four onboarding fields are collected: project_type, stack, kit_preset, and hooks_optin. Returns state='manifest_preview' with the install confirmation prompt.
Run Self Test
Run the claudekit self-test after the kit has been written. Advances state from kit_written to self_test_passed. The self-test is a lightweight post-install sanity check. Returns state='self_test_passed' + next_tool='complete_onboarding'.
Complete Onboarding
Complete the onboarding flow after the self-test has passed. Advances state from self_test_passed to first_session_ready (terminal). Returns state='first_session_ready' — onboarding is done.
Install Kit
Install the claudekit kit files into the project by writing them server-side via installer.write(). Call with confirm=False first to preview the manifest; then call again with confirm=True after the user confirms. State transitions from ready_to_install → installed ONLY on confirm=True.
Report File Contents
Report signal verification outcomes back to the server after reading project files. Call after detect_stack returns signals_to_verify.
Record Audit Result
Persist structured audit findings from a /claudekit:audit run. Call after summarizing findings so they appear in the personal-area dashboard.
Update Kit
Compute the diff between the current kit install and the latest version. Returns files_added, files_modified, files_removed for preview before writing.
Confirm Kit Update
Record a confirmed kit update after the user approves the diff. Pass the latest_manifest_id from the update_kit response.
Get Install Summary
Return the most recent install metadata for a session: detected stack, kit_version, manifest_id, and installed_at. Used by /claudekit:setup-deploy.
Get Stack Decision
Look up the stack playbook for a proposed technology choice and return a structured verdict: approved, discouraged, rejected, or no_playbook. Use this before implementing a new library or pattern to confirm it aligns with the project's approved stack decisions.
Get your MCP into directories
A working endpoint is step one. Directory coverage is the coordinated launch across ChatGPT, Claude, Cursor, the MCP Registry, and community indexes.
Directory coverage for brandsClaude Code ships dangerously by default. Claudekit gives it built-in protection, project memory, and specialist subagents — so it can't make big mistakes silently, doesn't forget what you decided yesterday, and says done when it actually is.
Use the MCP endpoint listed on this page in your MCP client configuration. One-click install pills support Claude, Cursor, VS Code, and other hosts. Copy the remote MCP URL if your client needs a manual entry.
MCPBundles probed 27 tools on the live server. The tool list on this page reflects what was discovered at the last refresh — connect your client to see the full set available to your session.
No provider sign-in was required during MCPBundles' probe. Your client may still need MCPBundles credentials depending on how you connect.
Operate Claudekit — Claude Code with a senior developer watching ove? Verify ownership to take over this directory entry.
This server appears in the MCPBundles directory. Verify you operate it to take over the listing — name, description, logo, contact email, and skill content. We email a 6-digit code to a maintainer address your server publishes in /.well-known/security.txt or /.well-known/mcpbundles.json. Free, takes about a minute.