Overview
Synapsor is an agent access-control layer for existing Postgres/MySQL data and production write proposals. Start with your existing Postgres or MySQL database: connect selected tables or views, expose approved tools to your agent framework, and use Synapsor to record evidence, stage proposed writes, require approval, and replay what happened.
Use Synapsor when an agent needs customer, account, order, invoice, ticket, CRM, or workflow data but should not receive raw database credentials, arbitrary SQL access, or direct production write authority.
Practically, the first use case is safer agent access to the database you already run. Technically, Synapsor is a hosted agent-native DBMS/runtime that owns evidence, proposal, branch, workflow, settlement, and replay state.
The V2 hosted runtime is controlled-beta, single-node Synapsor with a cloud control plane for projects, databases, API keys, billing, backups, and restore verification.
HTTP routes such as /v1/sql and /v1/control are the current stable API namespace for the V2 controlled-beta product. The /v1 prefix is not a V1 product label.
The production wedge is trust inside the database: hidden session bindings, evidence bundles, run graphs, write proposals, branches, policies, and replayable runs.
Developer notes
- Start with a database-scoped API key for application traffic.
- Keep tenant_id and principal in session context, not user-editable SQL.
- Use proposal mode for agent writes that require review.
- Use evidence bundles for answers that need auditability.