Skip to main content

Connectors

Every data source your agent can reach — scoped connections, one audit trail. Three ways to connect.

A connector is a data source your agent can reach through Hebrah. Each one is a registered target with a scope grammar: your agent opens a connection, names the tables and time period it needs, and queries through it — never seeing the underlying credential.

The catalog ships 12 demo targets across 6 packs, all healthy and queryable from a $1 trial — plus 8 live account connectors: connect your own Stripe, Postgres, Notion, Slack, Airtable, Linear, SFTP, or REST endpoints and agents reach the real data through the same scoped model. Run hebrah discover at any time to list what your agent can see.

Available connectors

Developer

Local and HTTP sandboxes for research agents and dev tooling.

Target Connector Inside Example scopes
Public REST

Synthetic JSONPlaceholder-style endpoints for research agents.

http_restposts, users, commentsposts:read,period:2025-Q4,tags:ai
Local research DB

10K rows of synthetic research data with tags.

sqlitearticles, sourcestable:articles,period:2025-Q4
Postgres (your database)live

Live Postgres data via a read-only connection URI you provide — declare the tables your agents may read.

postgresyour tablestable:orders,period:2026-09
HTTP REST (your endpoint)live

Query any JSON REST API with a bearer token — map logical tables to URL paths in the credential.

http_restyour logical tablesposts:read,tags:ai

Audit

General-ledger evidence, scoped to account ranges and periods.

Target Connector Inside Example scopes
Postgres audit ledger

Read-only general-ledger evidence for an audit engagement.

postgresglgl:read,period:2025-Q4,accounts:1000-3999

Fintech

Settlements, billing, and payment data for reconciliation.

Target Connector Inside Example scopes
SFTP settlements

Read-only settlement data for reconciliation workflows.

sftpsettlementssettlements:read,period:2025-11,merchant:<id>
Stripe billing (demo)

Synthetic Stripe API mirror — customers, charges, subscriptions, invoices, payouts.

stripecustomers, charges, subscriptions, invoices, payoutscharges:read,period:2025-11
SFTP (your file server)live

Settlement/batch CSV files from your SFTP server via a platform-held credential.

sftpsettlementstable:settlements,period:2026-08,merchant:M001
Stripe (your account)live

Live Stripe data via a read-only restricted key connected in the dashboard.

stripecustomers, charges, subscriptions, invoices, payoutscustomers:read,invoices:read,period:2026-09

Healthcare

Synthetic FHIR R4 — patient, encounter, observation. No PHI on the control plane.

Target Connector Inside Example scopes
Synthetic EHR (FHIR R4)

Synthetic Patient / Encounter / Observation resources for clinical agents.

fhirPatient, Encounter, Observation, Condition, MedicationRequestPatient:read,Encounter:read,period:2025-Q4

Workspace

Notion, Airtable, Slack, Linear, Google Drive — where the work happens.

Target Connector Inside Example scopes
Notion (your workspace)live

Live Notion databases and pages via your integration token — grant the integration to the databases agents should read.

notiondatabases, pagespages:read,database:<db-id>
Airtable (your bases)live

Live Airtable tables and records via a personal access token.

airtabletables, recordsrecords:read,table:Customers
Linear (your workspace)live

Live Linear issues, projects, teams, and users via an API key — WHERE filters map to Linear’s filter schema.

linearissues, projects, teams, usersissues:read,state:In Progress,team:ENG
Slack (your workspace)live

Live Slack channels, messages, and members via a bot token — invite the bot to the channels agents should read.

slackconversations, messages, usersmessages:read,channel:<channel-id>
Notion workspace

Synthetic Notion workspace with engineering + sales databases.

notiondatabases, pages, db-tasks, db-clientspages:read,database:<id>
Airtable pipeline

Synthetic Airtable bases with CRM-style customer + deal data.

airtabletables, records, Customers, Dealsrecords:read,base:<id>,table:Deals
Slack workspace

Synthetic Slack workspace with channels, messages, and member directory.

slackconversations, messages, usersmessages:read,channel:<id>,period:2025-11
Linear tracker

Synthetic Linear workspace with team-scoped issues, projects, and members.

linearissues, projects, teams, usersissues:read,team:<key>,period:2025-11
Google Drive shared

Synthetic Google Drive with strategy docs, runbooks, and marketing assets.

google_drivefiles, folders, permissionsfiles:read,folder:<id>

Manufacturing

Plant and line telemetry over MQTT.

Target Connector Inside Example scopes
Factory telemetry

Synthetic MQTT broker with plant / line / metric topics.

mqttmessagestopic:<glob>,plant:<code>

How to connect

Three ways in — same registry, same scopes, same audit. Open a connection, query it, export the evidence. The agent never receives a credential at any step.

1. Terminal — CLI

hebrah discover                                  # list every target + pack
hebrah inspect demo-postgres-audit               # scope grammar, tables, tiers
hebrah connect demo-postgres-audit \
  --scopes gl:read,period:2025-Q4 --ttl 3600     # scoped, TTL-bound connection
hebrah query <conn> --sql "SELECT * FROM gl LIMIT 10"
hebrah audit <conn> --verify                     # hash chain exits clean

2. Hosted MCP

Point any MCP host at https://mcp.hebrah.com/mcp with an agent key (hb_conn_…) and use the connection tools.

discover_data_sources            # targets + required scopes
connect_to_data_source           # target, scopes, tier, ttl
query_data_source                # rows, metering, audit event id
get_data_source_audit            # hash-chained events
revoke_data_source_connection    # instant, per connection

3. HTTP — headless

POST /v1/connections
{"target_id":"demo-postgres-audit",
 "scopes":["gl:read","period:2025-Q4"],
 "tier":"container","ttl_seconds":3600}
# → 201 {"connectionId":"…","audit_genesis_hash":"…"}

POST /v1/connections/{id}/query   {"sql":"SELECT * FROM gl LIMIT 10"}
GET  /v1/connections/{id}/audit
GET  /v1/audit/verify?connection_id={id}

Production targets work the same way: connect the credential in the dashboard (Settings → Credentials — a read-only Stripe restricted key, a connection URI, a Notion integration token, a Slack bot token…), and the agent discovers it as a target — the credential itself stays with Hebrah, scoped read-only.

Simple use cases

Research digest

A research agent pulls this week’s articles and tagged posts to draft a market digest, then exports the audit trail as its receipt.

Sources: Local research DB + Public REST

connect demo-sqlite-research --scopes table:articles,period:2026-09,tags:ai

Payment reconciliation

A finance agent joins general-ledger entries with Stripe charges and SFTP settlement files — scoped to one period — and leaves one hash-chained trail across all three sources.

Sources: Postgres audit ledger + Stripe billing (demo) + SFTP settlements

connect demo-postgres-audit --scopes gl:read,period:2025-Q4

Weekly ops standup

An ops agent reads last week’s Linear issues and Slack channels to draft a status update — scoped per team, channel, and month.

Sources: Linear tracker + Slack workspace + Notion workspace

connect demo-linear-tracker --scopes issues:read,team:eng,period:2025-11

Content brief

A marketing agent compiles the latest docs from a Drive folder and the client database in Notion into a content brief — every file and row it read is logged.

Sources: Google Drive shared + Notion workspace

connect demo-google-drive-shared --scopes files:read,folder:<id>

Get started

Start with $1 in free credit

Give your agent /SKILL.md and it onboards itself — 90 seconds, no card required.