InterLock by Context Data
Verified Oct 2, 2026 ActiveA self-hosted proxy that checks role, redacts fields and logs every request an AI agent makes against databases, files and APIs.
Listed on EffectHub ·
Pricing
- Pricing model
- Contact sales
No pricing is published. The site is recruiting design partners and routes everything through a booking link; the product is described as public beta with a 1.0 release candidate in the docs.
Use tool ↗About InterLock by Context Data
InterLock sits between AI agents and the systems they read from, speaking MCP, the Postgres wire protocol and HTTP, so agents keep their tools and only swap a connection string. Each agent gets its own key; requests are checked against source roles (down to table and column for SQL), personal data is masked in results, risky writes such as an unbounded DELETE wait for human approval, and every decision lands in an append-only audit log. It runs inside the customer's own environment via Docker Compose or Helm. Context Data is also the older data-pipeline vendor behind this site, which has now pivoted to this product.
- API
- Yes
- Last verified
- Oct 2, 2026
Verified sources
- InterLock is a proxy that authorizes requests by role and policy, redacts sensitive fields, holds risky writes for approval and audits everythingcontextdata.ai ↗
- Documentation covers connectors, deployment and the v1 contracts, with a 1.0.0-rc.1 release pathinterlock.contextdata.dev ↗
- Legal terms name The Context Data Inc. as the site ownercontextdata.ai ↗
Is InterLock by Context Data legit?
We don’t label tools safe or unsafe. Below are the records we checked ourselves, each with the date we read it.
- Last checked
- Oct 2, 2026The site answered
- Operated by
- The Context Data Inc.Our research, card verified Oct 2, 2026
Pros & cons of InterLock by Context Data
A narrow, infrastructure-level answer to a real gap: tool-permission platforms decide which tools an agent may call, but not which rows come back. It suits teams about to point agents at production Postgres or MySQL who want deny-by-default access and an audit trail without touching agent code. It is early and sales-led, so expect conversations rather than a self-serve trial.
Pros
- Works on the wire protocols agents already use, so adoption is a connection-string change
- Deny-by-default roles down to table and column, with SQL parsed before it runs
- Risky writes are held for a human reviewer, with optional Slack notices
- Publishes a per-feature status page that says what is beta, planned or disabled
Cons
- Pricing and company details are not published; access is through design-partner talks
- Several connectors and features are still marked beta or planned
- Another service in the data path adds an operational dependency and a latency hop
Best for
- Platform and security teams putting agents in front of production databases
- Regulated environments that need column-level masking and an audit trail
- Teams already using MCP clients or psql-style tools who cannot rewrite their agents
Not for
- Individuals or small teams wanting a hosted, self-serve product with public pricing
- Anyone looking for prompt-level content moderation rather than data-access control
Watch out: The vendor's FAQ and older pages still describe its previous data-pipeline product, so check you are reading the InterLock docs rather than the legacy ETL material.