Feanor pulls scattered company knowledge out of email, Slack, tickets, docs and databases, structures it into a living map of how the business actually works, and turns it into an executable skills file your AI agents can run on.
Connect the systems where your knowledge already lives. Feanor reads, normalises and refreshes the picture as the business changes. No new place to write things down.
1 # Auto-generated by Feanor · last refresh 2 min ago 2 name: handle_customer_refund 3 owner: [email protected] 4 trust_score: 0.94 # 312 prior cases reviewed 5 6 when: 7 - intent: refund_request 8 - order.age_days <= 30 9 10 decide: 11 - if order.amount < 200 → auto_approve() 12 - if customer.tier == "enterprise" → route_to(@maria) 13 - else → draft_reply() + flag_for_review() 14 15 grounded_in: 16 - slack://#cs-policies/2025-Q4-refund-thresholds 17 - notion://Customer-Ops/Refund-Playbook-v7 18 - zendesk://macros/refund-standard
Every company runs on critical know-how scattered across email, Slack, tickets, wikis, databases and people's heads. The business works because humans vaguely remember where it lives. AI agents can't operate that way. They need a primitive they can read.
A skills file is more than a doc. It's how AI agents do the work the same way a tenured employee would, every time, and prove it.
Skills regenerate as policies, prices and people change. The version your agent runs is the version your team agreed on this morning.
Every decision a skill makes traces back to the Slack thread, ticket or doc it came from. Auditable for legal, debuggable for engineers.
Owners review and sign off on each skill before it goes live. Confidence scores tell agents when to act and when to escalate.
Every change to a skill produces a diff. Roll back, A/B test, branch by team. Your operating manual under source control.
Per-skill permissions, redaction of sensitive sources, separation of read access from action. SOC 2 path on day one.
Skills export as portable YAML and as adapters for Claude, GPT, Gemini, Llama, and any tool-using framework you already run.
This isn't a wiki bot. The brain emits executable behaviour. Refunds happen. Incidents get triaged. Pricing exceptions get decided.
Your knowledge never trains a foundation model. Single-tenant deployments and self-host options for regulated environments.
Track which skills run, how often, what they save. Tie the brain to outcomes the CFO actually cares about.
The biggest blocker to AI automation of companies is no longer the models. They got good, fast. The blocker is domain knowledge. Every company in the world is going to need a brain.
One console for the people who own the brain. See what skills exist, who signed off on them, where the agents are running them, and what the business is getting back.
Feanor is the missing layer between raw company data and reliable AI automation. A primitive every business will need.
"A company-wide search shows you the document. The brain tells the agent what to do. That distinction is the whole game right now."
"Models stopped being the bottleneck about a year ago. The reason agents still trip is they don't know what 'we' do. The brain fixes that."
"Your brain is your moat. We help you build it, keep it current, and run it safely. We don't keep it. You do."
Alex is the AI agent that reads inbound mail to Feanor. Write a real question (interest, pricing, technical, anything). You will get a real reply from [email protected] in under sixty seconds.