AeonBrain

Open-source CLI · Apache 2.0 · signed + audit-logged from Day 1

Your AI always runs on current policy.

Upload your SOPs, policies, and job aids as PDF, Word, or Markdown. Aeon extracts signed, executable skills your AI can follow - and when your docs change, your skills update automatically. Free with your own Anthropic key.

$ npx @aeonbrain/cli login
$ aeon set-key sk-ant-…
$ aeon connect markdown ./docs
✓ 47 files indexed
$ aeon extract
⏳ drafting skills with Claude…
✓ 12 skills drafted
# review and approve in the curator UI
$ aeon mcp-url
https://aeonbrain.airiskmanagement.ca/mcp/aeon_T2bEF…/sse
# paste this URL into Claude.ai → Add custom connector
# (leave OAuth blank). Requires a paid Claude plan.

No credit card. CLI is open-source. Schema is Apache 2.0.

From a doc to a signed skill

This is the actual output: a typed, governed, cryptographically-signed skill. No signup needed to see what you get.

Your doc · refund-policy.md
# Refund policy

For orders over $100, look up the order
in Stripe, confirm the customer tier, and
if Gold, approve up to a full refund.
Anything above $500 needs a manager's
sign-off. Log every refund to #finance.
Aeon skill · customer-refund-process
signed ✓
steps:
  1. lookup         order in Stripe
  2. lookup         customer tier
  3. user-input     manager approval (>$500)
  4. external-call  POST refund
  5. llm            draft #finance note
governance:
  dataClasses:      [PII, financial]
  requiresApproval: true
ed25519: 3f8a…b21c · tenant key · verified at the runtime gate
Append-only audit log · every call, immutable
skill.executedvia: mcp (Claude)order #4471 · $2402026-06-08 14:21Z

One depository, every AI runtime

Upload any format

Drop in PDFs, Word docs, or Markdown. Aeon reads your SOPs, policies, and job aids and extracts the workflows inside. No reformatting, no copy-paste into a chatbot.

Always current

When policy changes, re-upload. Aeon detects what changed, re-extracts only the affected skills, and flags them for curator review before republishing. Your AI never runs on stale procedure.

Signed, vendor-neutral skills

Every published skill is ed25519-signed by your tenant key. The same skill runs on Claude (MCP), Microsoft Copilot (OpenAPI), and ChatGPT (Custom GPT actions). Write it once, run it anywhere.

Append-only audit log

Every skill call AND document query writes an immutable row with full provenance. View it on any plan; CSV export (for SOC 2 / EU AI Act compliance) unlocks on Pro and above.

Find gaps and contradictions

Aeon audits your whole corpus and flags where two docs disagree, where a procedure is only half-documented, and which skills you could build but haven't yet. High-confidence findings only.

A living wiki, auto-compiled

Every doc gets a 3-bullet summary, a domain tag, and backlinks to related docs - a browsable, always-current map of everything your company knows. Regenerate it any time.

Simple pricing

Free with your own Anthropic key. Upgrade for managed LLM, auto-update when docs change, and the compliance audit export.

Free

custom
  • BYOK - your Anthropic key, your bill
  • Cryptographic signing on every skill
  • MCP for Claude.ai + open-source CLI
  • Audit log (view in-app)

Pro

custom
  • Managed LLM - no API key, we run the model
  • Audit-log CSV export (the compliance report)
  • No "Powered by Aeon" footer
  • Email support

Team

custom
  • Everything in Pro, for your whole team
  • 5 seats - multi-user collaboration
  • Auto-update skills when docs change
  • OpenAPI for Microsoft Copilot Studio

Governance

custom
  • Unlimited skills + daily auto-sync
  • Audit CSV export (compliance evidence)
  • Priority support + a setup call
  • SSO / SAML on the roadmap

Open by default at the layers that matter

@aeonbrain/cli and @aeonbrain/skill-schema are Apache 2.0 - fork them, audit them, extend them. The skill schema is published as a wire format so the same skill file works across Claude, Copilot, and (V2) ChatGPT. The hosted platform around it is proprietary.