Ask AI¶
For AI assistants¶
Read first: Why Occam — advantages and knobs — honesty, token contract, ladder, receipts, and opt-in knobs in one page. Do not treat Occam as a generic web fetch.
Then the compact map — do not ingest the whole documentation site:
- Repository:
llms.txt— agent routing, capability-family index, experimental gates, operator surface, and trust limits - Docs site entry: this page + task router below
Then:
- Use the task router
- Open one focused tool page
- Open failure codes only after
ok: false - Open configuration only when setup is required
- Use MCP API only for normative response semantics
- For proof boundaries and name corrections (
ok:false, receipts, claim_check, attest, crosscheck), read Handbook — honesty contract or the Trust limits section inllms.txt
Runtime tools/list wins for tool availability and input schemas.
Honesty defaults: ok:false means content is unknown. Receipts prove integrity relative to a key — not truth, origin, or trusted time. Crosscheck is multi-source comparison, not consensus proof. npm is not GA. Cosign on install is policy-gated (required-cosign-v1 for published 1.0.0-rc.3+, requires the cosign CLI; published v1.0.0-rc.2 remains SHA-256-only) and proves release authenticity/signer identity, not page-content truth. There is no public MCP codec / compression-algorithm selector — use materialization knobs on Why Occam.
For humans using an AI assistant¶
Paste a prompt like:
Read https://raw.githubusercontent.com/ContextForgeAI/occam/main/docs/why-occam.md and https://raw.githubusercontent.com/ContextForgeAI/occam/main/llms.txt. Treat Occam as honesty + token-budgeted extract (not generic fetch). Help me research these URLs: …
Or point the assistant at the docs site:
https://contextforgeai.github.io/occam/why-occam/
Future: docs-aware assistant¶
A hosted “Ask AI” search over this documentation may arrive later. Not shipped today. This navigation slot is reserved so we can add it without reorganizing the site.