README.md |
the pitch, quickstart, one contained tool call, headline numbers, honest limits |
Entry point for humans; numbers bound in benchmarks/claims.yaml |
ee/README.md |
what AgentFox Enterprise adds (compliance, evidence, monitoring, access checks), its licence, how it plugs into the open core |
Source-available, BSL 1.1 (ee/LICENSE); the open core never imports it |
ARCHITECTURE.md |
the domain model, one tool call traced through auto() and the gateway, the code map, invariants, where to start for a change |
Start here before changing code; names functions, not line numbers |
docs/product-tour.md |
every integration surface (auto, gateway, LangGraph, OpenAI Agents SDK, TypeScript, MCP, Claude Code hooks), commands grouped by task, demo output, self-hosting |
Moved out of the README; the website guides cover the same ground |
docs/design/PRD.md |
why a feature exists, requirement IDs, roadmap, non-goals |
1,200 lines — jump to §5 pillars, §6 integrations, §9 roadmap, §10.3 non-goals, §12 shipped since |
docs/architecture/high-level-design.md |
architecture, request path, deployment topology |
Derived from code; code wins |
docs/architecture/low-level-design.md |
which module implements what, class internals, CLI tree |
Derived from code; code wins |
docs/design/traceability.md |
requirement → module → control → test |
Hand-maintained |
docs/design/failure-modes.md |
how deployed agents fail (F1–F9) and which are covered |
Living scorecard |
docs/design/gap-analysis.md |
enterprise-readiness gaps, Tier 0–3 |
Canonical gap register |
docs/architecture/judgment-tiers.md |
what each optional judgment tier (Jev, hosted LLM, local LLM) may decide, and the routing table |
Read before enabling a tier |
docs/architecture/judgment-egress.md |
what leaves the boundary when a hosted judgment tier is on; the PII block/redact/allow choice |
Read before enabling a hosted tier on regulated data |
docs/design/production-readiness-review.md |
gaps found re-checking HLD/LLD against the code, and what was fixed since |
Dated review; docs/design/gap-analysis.md + git log for current state |
docs/design/competitor-analysis.md |
market landscape and where competitors win; source for the public /compare page |
Positioning; cite primary benchmark docs for numbers |
docs/design/competitor-failproof-docs-and-product.md |
Failproof AI's docs structure (IA, page template, site features) and product, against ours; ranked docs and product gaps |
Dated review (2026-10-07); docs fixes it lists are done, product gaps are proposals |
docs/design/architecture-differentiation.md |
the architecture-vs-competitor pitch: where each vendor sits and what it can see (sourced from their docs and code), primitive → benchmark → claim id, objection handling, proof gaps |
Sales-facing; numbers bound in benchmarks/claims.yaml; competitor facts labelled [code]/[docs]/[ND] with access date |
docs/evaluation/judgment/*.md |
the judgment-tier study: results (numbers bound in benchmarks/claims.yaml), vs-shipping, cascade, safepyramid |
Quote numbers from results.md only |
docs/evaluation/responding-to-the-critique.md |
answering the "AI security is bullshit" argument: what we accept, what we dispute, what we changed |
Sales-facing; keep claims checkable |
docs/design/control-catalog.md |
NOM-* controls and their framework mappings |
All mappings DRAFT; source of truth is src/agentfox/packs/compliance/catalog/controls/controls.yaml |
docs/design/compliance-crosswalk-review.md |
our draft framework mappings cross-checked against a second published mapping: gaps both ways, disagreements, a recommended action per row |
Dated review at a pinned upstream commit; lists every catalog change it made |
docs/architecture/threat-model.md |
threats to customers' agents and to the product |
|
docs/architecture/data-model.md |
the persisted tables and their columns |
Derived from src/agentfox/core/models/; code wins |
docs/architecture/api-spec.md |
every HTTP route, authentication, the inline enforcement routes |
Route tables generated by scripts/gen/api_routes.py --write; prose hand-written |
docs/design/structure-proposal.md |
the agreed code-restructure plan: target layout, harness adapters, capability packs, phases and progress |
Decided 2026-10-06; check its Progress section before restructuring |
docs/adr/*.md |
why the repository is shaped the way it is: one decision per record, with the alternatives that lost |
docs/adr/README.md indexes them; a record is superseded, never rewritten |
docs/design/oss-register.md |
why an OSS project is (not) used, licences |
Check before adding a dependency |
docs/README.md |
the index of docs/: what each document is for and who it is for |
Start here when you do not know which document answers a question |
docs/getting-started.md |
a linear first hour for a new user, ending with their own agent governed |
Written for a newcomer, not for an agent; each step says what it proves |
docs/deployment/systemd.md |
running the Python gateway as a reboot-persistent systemd user service, with separate database, evidence and signing-key storage |
Self-hosters on Linux |
docs/deployment/key-rotation.md |
rotating the token encryption and audit signing keys: *_PREVIOUS settings, agentfox admin keys rotate/status, the automatic keys.rotate job, what to do when rotation cannot finish |
Anyone running a deployment |
docs/evaluation/evidence-standards.md |
how to read any number in this repo, and the limits that apply before it |
Read before quoting a benchmark |
benchmarks/README.md |
which benchmark proves which claim |
Index; quote numbers from the linked methodology only |
THIRD_PARTY_NOTICES.md |
attribution obligations |
Update when adding a dependency |