Run agents on knowledge they're allowed to see
EigenSearch lets your agents query permission-filtered company knowledge and call your own extensions, from eigenpal or any workflow engine.
A run, step by step
The agent answers a due-diligence question using only documents the requester can open, then calls your internal risk extensions.
Expose your internal tools as extensions
Wrap the tools your team already built (risk models, KYC lookups, contract review) as extensions. Agents call them during a run, scoped to each user's permissions.
Works with any workflow engine
eigenpal is the default backend. The same interface works with LangGraph, Temporal, or your own engine. No lock-in.
- One contractRetrieval and identity work the same way for every engine.
- Swap freelyMove from eigenpal to LangGraph, Temporal, or your own engine.
- No lock-inYour agents call EigenSearch the same way, whatever runs them.
Knowledge, security, and orchestration in one place
One system for permission-filtered retrieval, identity resolution, and the tool calls your agents make.
Scoped to the user
Every lookup runs as the person asking. Agents never surface a document they couldn't open themselves.
Document-level security
Permissions are resolved at query time from your existing identity and groups, not copied or cached.
Orchestration that acts
Agents don't just answer. They call your extensions and workflows to move the work forward.
See how EigenSearch answers with sources
Walk through how EigenSearch connects to your internal sources, respects permissions, and returns cited answers your teams can verify.
Part of the