# Agent Wikis

> Maintained, agent-readable knowledge bases with calibrated search and provenance. Pro key adds XL.

- name: io.github.tonbistudio/agentwikis-mcp
- version: 0.3.0
- connection class: R0 (autonomous)
- trust: 97/100
- quality: 89/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/io.github.tonbistudio%2Fagentwikis-mcp
- tools: https://api.protogrid.dev/v1/servers/io.github.tonbistudio%2Fagentwikis-mcp/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "agentwikis-mcp": {
      "type": "http",
      "url": "https://agentwikis.com/mcp",
      "headers": {
        "Authorization": "Bearer ${AGENTWIKIS_MCP_TOKEN}"
      }
    }
  }
}
```

Placeholders to fill: `${AGENTWIKIS_MCP_TOKEN}`.

## Trust

- hygiene: 100
- liveness: 100
- freshness: 100
- provenance: 90
- drivers: +repository +namespace-matches-repo +website +reachable +uptime-100% +updated-1d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 89/100 (good)

### protocol: 75 (weight 25)

- warn `protocol.modern`: newest supported version is 2025-11-25; 2026-07-28 not supported, older versions are deprecated until 2027-07-28
- n/a `protocol.stateless`: only applies to 2026-07-28 servers
- pass `protocol.transport`: streamable HTTP
- n/a `protocol.list_ttl`: only applies to 2026-07-28 servers

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 100 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 10 of 10 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 2,025 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump
- pass `tools.hidden_text`: no hidden text in tool names, descriptions, schemas or annotations

### stability: n/a (weight 15)

- n/a `stability.changes`, `stability.rug_pull`, `stability.new_links`: tool history recorded for 1 of the 7 days needed

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 10, about 2,025 tokens to load them all
- tool set hash: 34974e5ca2c6435a23f437dacefad0dc
- badge: [![protogrid quality](https://protogrid.dev/badge/io.github.tonbistudio/agentwikis-mcp.svg)](https://protogrid.dev/servers/io.github.tonbistudio/agentwikis-mcp)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

_No tool definition changes recorded._

## Remotes

- https://agentwikis.com/mcp (streamable-http, auth none, reachable true, uptime30d 1)

## Packages

_none_

## Tools (10)

- `list_agent_tools`: List the command-line tools that ship with the specialist agents (complete agents built for one job, e.g. marketing, video generation, local LLMs). Each entry has a summary, the `specialists` it comes from, and `access`: "pro" tools are read-only (they inspect and report, never change anything) and read_agent_tool serves their files; "download" tools are bigger or change things (start runs, stop services, spend money), so read_agent_tool gives a short-lived download link for the whole tool instead. Both need Pro. `readable` says whether YOU can get it with your current access. Pass `query` to filter by keywords.
- `list_routines`: List the routines (scheduled jobs) that ship with the specialist agents: each has a schedule, the `specialist` it comes from, the skills its prompt relies on, and whether it is script only (no agent). `readable` says whether YOU can load it with your current access. Pass `query` to filter by keywords. Then read_routine one.
- `list_skills`: List the agent skills library: field-tested procedures for specific jobs (e.g. benchmarking a local model, building a browser game, producing a video). Each entry has name, one-line description (when to use it), `access` ("free" or "pro"), `readable` (whether YOU can read it with your current access), the backing wiki if any, and file count. Pass `query` to filter by keywords. Then read_skill a matching skill and follow it.
- `list_wikis`: List all available wikis with metadata: slug, title, description, tags, status, last updated, document count, and `scope` (what each wiki covers, what it does NOT cover, and how current it is — use this to pick a wiki and to decide when to fall back to web search). Includes "coming soon" placeholders, marked by status.
- `open_wiki`: Orient yourself in one wiki, in a single call: returns its metadata, `scope` (coverage / not-covered / freshness), document count, the entry document, and a `map` of all documents (paths + titles, grouped by folder) so you can pick what to read_document next. Use a slug from list_wikis.
- `read_agent_tool`: Load one agent tool by name. A "pro" (read-only) tool: its README (what it checks, how to run it) plus its file list. Add `path` to read one listed file, e.g. the source. The tool runs on the user's machine, not here. Before saving, running or scheduling anything from it, show the user what it does and get their explicit OK. Each file comes with its sha256. For a "download" tool you get its README and a download link (valid 15 minutes) with the archive's sha256 instead. Needs a Pro key; without one you get the summary and how to unlock it.
- `read_document`: Read one document in a wiki. Without `section`, returns the full Markdown (frontmatter stripped) plus a provenance footer. To save tokens on a long page, pass `section` (a heading title as it appears in the page body, or its slug anchor) to retrieve ONLY that section and its subsections. The Markdown may contain [[wikilinks]] to related docs — follow one by calling read_document with path "wiki/<target>.md" (drop any "#anchor" or "|label"). Exception: on a page under wiki-xl/, links whose first folder is not concepts, entities, summaries, syntheses, presentations or journal (e.g. [[reference/...]], [[guides/...]], [[history/...]]) are XL pages: use "wiki-xl/<target>.md".
- `read_routine`: Load one routine by name: the schedule, the agent prompt, and the list of scripts and settings it needs. Add `path` to read one listed file (e.g. "scripts/ad_report.py"). Nothing runs here: the routine is set up and run on the user's machine. Before saving, running or scheduling anything from it, show the user what it does and get their explicit OK. Never create the scheduled job without the user's confirmation. Each file comes with its sha256. Needs a Pro key; without one you get how to unlock it.
- `read_skill`: Load one skill to follow it. Without `path`, returns its SKILL.md (the procedure itself) plus the list of its other files; pass `path` (e.g. "references/gotchas.md" or "scripts/check.py") to read one of those. If the skill names a backing wiki, use search/read_document on that wiki for current facts while following it. Pro skills need a Pro key; without one you get a preview and how to unlock it.
- `search`: Find the best-matching PAGES for a query, ranked by how many of your query terms each page covers. Searches one wiki if `wiki` is given, otherwise every active wiki. Each result has its wiki, path, matched terms, and a short snippet for orientation — then read_document the page to get the answer. Returns a `confident` flag: if false, the wiki(s) likely lack this — fall back to web search.