Why are my OpenClaw MCP tools missing even though the server is configured?

The server may be registered, but tool discovery can still fail because the process cannot start, returns zero tools, is hidden by the active tool profile, or is filtered before the model request is built.

Should I reinstall OpenClaw when MCP tools do not show up?

Usually no. Confirm registration, static doctor output, probe results, tool profile, deny rules, gateway logs, and subagent scope before reinstalling.

Can subagents have fewer MCP tools than the parent agent?

Yes. If the parent can call a tool but the subagent cannot, check subagent tool inheritance, runtime scoping, and any profile or filtering rules applied to spawned agents.

Usage Tips

OpenClaw MCP tools missing? Diagnose discovery before reinstalling

Problem statement: you configured a Gmail, Outlook, GitHub, database, filesystem, or custom MCP server, but the agent says it does not have access to those tools. Sometimes the server appears in config. Sometimes direct probes work. The real failure is often between config registration, tool discovery, active profile injection, and subagent scope.

Fast answer

Confirm the server is registered, run static MCP status or doctor checks, probe the server only after static config passes, inspect gateway MCP logs, confirm the active tool profile is not minimal, check tools.deny and per-server toolFilter, then test from the parent agent before blaming the model or reinstalling OpenClaw.

Symptom map

Server missing from config or list: look for the wrong config path, a bad reload, or a profile that is not using the expected config.
Server listed but zero tools: check the command, PATH, working directory, OAuth completion, and whether the server returns a valid tools/list response.
Probe works but agent cannot call tools: inspect runtime tool injection, profile selection, deny lists, and include/exclude filters.
Parent can use the tool but subagent cannot: treat it as a subagent inheritance or scoping issue instead of a server startup issue.

Hosted-team checklist

Capture evidence first. Save OpenClaw version, install method, runtime profile, exact error phrase, and the MCP server name.
Run static checks first. OpenClaw MCP docs distinguish status/list/show/doctor checks that read config from probe checks that connect to the server.
Use probe after static config passes. A probe can prove process startup and tool listing, but it does not prove the agent runtime injected those tools into the model request.
Confirm the profile. Normal coding and messaging profiles can expose configured MCP tools, while minimal profiles hide them.
Check filters. tools.deny = ["bundle-mcp"] can disable bundled MCP exposure, and per-server toolFilter rules can include or exclude discovered tools.
Restart the gateway once. After a config change, capture logs, restart once, then run a single production-like prompt.

Limited managed setup experiment

Fix once. Stop recurring MCP tool discovery failures.

If this keeps coming back, you can either move the setup path into managed OpenClaw hosting or book the constrained launch package for one workspace. The experiment is deliberately scoped: one hosted instance, first-run configuration, channel/setup guidance where supported, one smoke test, and a handoff note.

$199 managed setup One hosted OpenClaw workspace, one 30-minute onboarding/debug session or equivalent async help, and a 7-day setup-specific follow-up.

Clear boundaries before work starts No custom development, enterprise/SRE support, unsupported self-hosting repair, or open-ended third-party debugging.

Includes hosted instance setup, first-run configuration, channel/setup guidance where supported, smoke test, and handoff note
Excludes unlimited support, custom workflow/code work, unsupported self-hosting repair, and third-party provider outages
Limited weekly slots keep the experiment operationally safe while setup time and lead quality are measured

If you would rather compare options first, review OpenClaw cloud hosting or see the best OpenClaw hosting options before deciding.

Book managed setup - $199 setup, limited weekly slots Move this setup to Lobsterland Keep self-hosted: apply fix checklist

OpenClaw import first screen in OpenClaw Setup dashboard (light theme) — 1) Paste import payload

OpenClaw import first screen in OpenClaw Setup dashboard (dark theme) — 1) Paste import payload

OpenClaw import completed screen in OpenClaw Setup dashboard (light theme) — 2) Review and launch

OpenClaw import completed screen in OpenClaw Setup dashboard (dark theme) — 2) Review and launch

Common server-side causes

If a server is listed but has no tools, the most common causes are mundane: the command cannot be found from the gateway process, uvx or another launcher is absent from the daemon PATH, the working directory is wrong, required environment variables are missing, OAuth has not completed, or the server returns a malformed tools list. Running the MCP server manually from the same user and environment can shorten the search.

When managed hosting helps

Managed hosting does not make bad MCP servers magically correct. It does make the boring parts stable: secret paths, gateway restart behavior, log capture, runtime profiles, and safe isolation for custom servers before they are attached to production agents.

CTA

If MCP tool discovery is eating a team onboarding session, move the instance to Lobsterland and keep gateway logs, persistent config, and deployment-safe MCP changes in one hosted runtime.

See managed OpenClaw hosting Open dashboard

Verification checklist

The server appears in the intended config and static MCP checks show no command, working-directory, TLS, disabled-server, or OAuth warnings.
A probe lists the expected tools from the server.
The active profile is coding or messaging, not minimal.
No deny-list or toolFilter rule removes the expected tools.
A parent-agent prompt can call one expected tool successfully.
A subagent prompt can call the same tool when subagent inheritance is expected.