Secure data access
for AI agents.
Gives your coding agents scoped access to private data like Postgres, SFTP, FHIR, Notion, and more with a verifiable audit trail on every query. Start with $1 in free credit, no card required.
Give your agents superpowers without exposing the keys to the kingdom
Healthcare
Give clinical agents scoped FHIR access for prior-auth checks, census monitoring, and chart review — synthetic data today, hash-chained audit evidence, and no PHI on the control plane.
Workspace automation
Ops and IT agents pull context from Notion, Slack, Linear, and Airtable through scoped connections — no long-lived API tokens in scripts, and every access is logged and revocable.
Research
Competitive-intel and research agents query the databases and files behind their answers — scoped to tables and time periods, metered per query, with a verifiable trail of exactly what was read.
Marketing
Reporting agents compile campaign performance from warehouses, Stripe exports, and SFTP drops — TTL-bounded sessions, per-query metering, and shareable evidence of what the agent saw.
Outbound-only relay
How the relay works
The customer runs one command — hebrah relay enroll — and the relay registers with the control plane via JWT, receives an mTLS certificate from step-ca, and opens a single outbound tunnel. No inbound firewall holes, ever.
Agent queries route through the control plane to the relay, which speaks the Postgres wire protocol locally with scopes enforced before any statement runs. Hash-chained heartbeats stream back as evidence, certificates renew themselves, and revoking a relay_id cuts access instantly.
Explore relay map nodes
hebrah-relay
Customer-side relay binary. One outbound tunnel, one scoped identity — private data stays reachable without opening inbound access.
Explore agent map nodes
Your agent
One scoped agent key opens connections to every source it needs — Postgres, Stripe, Notion, Google Drive, Slack — each with its own scopes, TTL, and audit chain.
20 connector targets across 6 packs
One agent, every data source
Your agent signs up once, then opens a scoped connection per source — Postgres for the ledger, Stripe for billing, Notion, Google Drive, and Slack for the workspace. Each connection names its tables and time period, carries a TTL, and never exposes a credential.
Every query across every connection appends to one hash-chained audit trail — a single verifiable evidence stream for everything the agent saw. Revoke any connection instantly without touching the rest.
Agents changed the threat model
A script holds one token for one service. An agent holds a working context across many — it copies secrets into prompts, follows instructions at machine speed, and moves far more data than any human session. Hand it a warehouse password and every query it runs is unscoped, unmetered, and invisible to your customer.
Hebrah fixes this at the connection layer: agents never hold a credential, access is scoped to tables and periods with a TTL, every query appends to a hash-chained audit trail, and any connection revokes in one call. Because access is governed rather than blocked, agents still work at real data scale — millions of rows, across every source — metered per query, with exportable evidence your customers can verify.
Explore security map nodes
Control plane
Hebrah sits between agents and private data — credential resolution, scope rewrite, metering, and audit all happen server-side.
Works with your agent stack
Connect Cursor, Claude Code, Codex, and other MCP hosts to Hebrah in minutes.
Agent loop: sign up headless in 90 seconds, discover targets, open a scoped connection, query, and export the hash-chained audit as your evidence trail. Give your agent /SKILL.md to onboard itself, or start in the dashboard at localhost:3000.