home / for-agents
For agents
Everything on this site is consumable without a browser, without JavaScript, and without an account. Pick the interface that fits your harness. The Pro (XL) tier doesn't add a separate interface — one API key widens the ones below.
1. llms.txt — start here
The standard entry point. An index of every wiki with its description, declared scope (what it covers and what it deliberately doesn't), freshness, and a raw-Markdown link for every page:
curl https://agentwikis.com/llms.txt
- /llms.txt — the index (wikis, scope, every page)
- /llms-full.txt — the entire corpus in one fetch
- /index.json — the same registry as JSON (wikis + skills)
- /sitemap.xml · /robots.txt
2. Raw Markdown
Every page is served as its exact on-disk Markdown — YAML frontmatter included (type, tags, updated, sources) — at:
/raw/<wiki>/<path> e.g. /raw/vllm/wiki/index.md
Or just request a human URL with Accept: text/markdown — you'll be redirected to the raw file. HTML pages also declare <link rel="alternate" type="text/markdown">. Raw responses carry ETag + Last-Modified, so conditional GETs make freshness polling nearly free.
Pro (XL) pages live under /raw/<wiki>/wiki-xl/<path> and answer 402 without a key; send Authorization: Bearer <api key>. Every wiki's full page list — both tiers, each entry flagged gated — is public at /wiki/<wiki>/index.json.
3. MCP server
The same wikis over the Model Context Protocol — four read-only tools shaped for an agent's loop: list_wikis (discover + scope), open_wiki (orient: metadata + document map), search (ranked pages + a calibrated confident flag), read_document (full page or a single section, with provenance). No API key or account needed for the free tier:
# Claude Code:
claude mcp add agentwikis -- npx -y agentwikis-mcp
# any MCP client:
{ "mcpServers": { "agentwikis": { "command": "npx", "args": ["-y", "agentwikis-mcp"] } } }
Pro (XL) over MCP. Set AGENTWIKIS_API_KEY to your Pro key and all three read tools widen to the XL tier — open_wiki returns the XL pages in its map (marked gated), search ranks across both tiers with confident calibrated over the full corpus, and read_document returns XL pages instead of 402. Requires agentwikis-mcp 0.2.0+:
{ "mcpServers": { "agentwikis": {
"command": "npx", "args": ["-y", "agentwikis-mcp"],
"env": { "AGENTWIKIS_API_KEY": "aw_..." } } } }
Without a key those tools serve the free tier and say so: open_wiki returns an xl field with the Pro page count, and search reports how many XL pages in scope it did not search — so a confident hit is never mistaken for "the best the wiki has".
Source: github.com/tonbistudio/agentwikis-mcp (MIT). The same server can be self-hosted against your own instance via AGENTWIKIS_URL. A hosted (remote-transport) endpoint is still planned — email support@agentwikis.com to hear when.
4. Skills (Pro)
Curated agent skills developed over months of real Hermes Agent work — not evals — and refined for release. Each is a thin SKILL.md discipline (plus its references and scripts) that pulls its facts from a maintained wiki over MCP. Discover them in /index.json (skills[]) and /llms.txt; browse at /skills. Full content and an installable bundle are included with Pro:
# browse the tree (HTML): /skills/<name>
# install the whole skill (Pro — send your key):
curl -fL -H "Authorization: Bearer $AW_KEY" \
https://agentwikis.com/skills/<name>/bundle.tar.gz -o <name>.tar.gz
tar -xzf <name>.tar.gz -C ~/.hermes/skills/custom/
5. Trust semantics
- scope — every wiki declares covers / notCovered / currentAs (freshness is derived from the wiki itself, never hand-typed). If your question is out of scope, use the web instead.
- confident — MCP search returns a calibrated flag; when
false, the wiki likely lacks the answer (measured ~94% correct abstention). Fall back to web search rather than trusting a weak match. The flag is always calibrated over the corpus you can read: without a Pro key that's the free tier, and the response'sxl.notSearchedtells you how many deeper pages were held back — soconfident: truemeans "the best page available to you", not necessarily the best page that exists. - provenance — pages cite their sources in frontmatter; MCP reads append a citable reference footer.
