mcptoon
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| prompts | {} |
| resources | {} |
| extensions | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| mcptoon_manifestA | List the upstream tools this gateway currently exposes, grouped by server. Returns names by default; that compact view is what an agent should read instead of pulling every full schema. Pass include_descriptions for a one-line summary per tool. |
| mcptoon_serversA | Report the MCP servers configured behind this gateway: transport, tool count and whether their tools loaded. Use it to see what is actually reachable before calling a tool, and to spot a server that configured but published nothing. |
| mcptoon_healthA | Self-check of this gateway process: version, config path, servers configured, tools indexed and uptime. Read this first when a tool you expect is missing; it distinguishes an empty config from a server that failed to load. |
| mcptoon_usageA | How much this gateway has saved: tools compressed, tokens saved and the session call count. Call this when you want to tell the user what mcptoon did. Report these numbers verbatim; never estimate them yourself. |
| mcptoon_skillsA | List the skills mcptoon indexes on this machine: slug and one-line description per skill. Read it once to learn what skills exist, then resolve the one you need with mcptoon_resolve_skills. Offline and read-only; pass include_aliases to also see alias cards. |
| mcptoon_resolve_skillsA | Return the skills that best match a task description, best first. Call this with the user's request before acting, then read the chosen skill's SKILL.md. Offline BM25 ranking, no LLM. |
| mcptoon_inspectA | Show one upstream tool's full input schema — the parameters to call it with. Read-only; nothing runs. Use it after mcptoon_manifest names a tool. |
| mcptoon_callA | Run any upstream tool by its namespaced name (server_tool). Under the default compact exposure the tool list is withheld, so this is how you run one found via mcptoon_manifest. Arguments are validated first. |
| mcptoon_retrieveA | Get back the full, uncompressed text of a tool result mcptoon compressed. When a result ends with a line naming a handle, call this if the summary lacks a detail you need. |
| mcptoon_kb_searchA | Return the PASSAGES in already-indexed documents that match a query, not the whole file. Use it instead of reading a large document whole when you need a few lines from it. Index a document first with mcptoon_kb_index. |
| mcptoon_kb_indexA | Chunk a local file or a URL into the knowledge base so mcptoon_kb_search can find passages in it. Indexing is what keeps the document out of your context window while staying searchable. |
| mcptoon_sessionA | Summarise what has routed through this gateway in this project: tools used, files written, errors, git operations. Call it after a context compaction to catch up on where you were. It is on-demand, not automatic — the gateway cannot inject into your context itself. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 12 tools
Each tool has a describable job and the descriptions cross-reference each other (manifest→inspect→call, skills→resolve_skills, index→search→retrieve), which makes the intended chains clear. The weak spot is the meta/status cluster: health, servers, session and usage are all 'report something about the gateway', and an agent may briefly hesitate between health and servers or between session and usage.
Every tool shares the same mcptoon_ namespace prefix and snake_case, so the set reads as one family. The trailing segment mixes verbs (call, inspect, retrieve, resolve_skills) with bare nouns (manifest, session, health, servers, skills, usage), so there is no strict verb_noun grammar, but it is predictable and readable.
Twelve tools is squarely in the sweet spot for a gateway. Each one maps to a distinct need — discover, introspect, invoke, decompress, index, search, skills, and observability — and none looks like filler.
The invoke path (servers→manifest→inspect→call→retrieve) and the skills path are fully covered, with health and usage for diagnostics. The knowledge-base path is the gap: you can index and search, but there is no way to list what is indexed, delete or re-index a stale document, so an agent can create state it cannot clean up.