graphjin.com

What it is

GraphJin positions itself as a “compiler connecting AI with organizational systems.” Its core idea is to expose databases, remote APIs, files, object storage, source code, and workflows through a unified, governed GraphQL + MCP surface. It emphasizes that it is not about piling up wrappers or resolvers, but about enabling Agents to understand systems and perform actions through observable evidence such as catalog, security, validate/preview, and audit.

Core capabilities

Functionally, GraphJin can automatically discover tables, columns, relationships, source-code metadata, and configured capabilities, then compile nested GraphQL into optimized database workloads. The goal is to reduce N+1 issues, resolver bloat, and ORM middleware layers. It supports PostgreSQL, MySQL, MariaDB, MongoDB, SQLite, SQL Server, Oracle, Snowflake, Redshift, BigQuery, Cassandra, Aurora, Cloud SQL, and more, while also covering HTTP APIs, S3/GCS/R2, files, source-code search, and OpenAPI joins. For AI use cases, it provides MCP and can integrate with Claude, Codex, Cursor, or custom MCP hosts; development can use stdio, while team access can run through an HTTP+SSE endpoint.

Governance and security

Its main differentiator is governance: a single configuration controls what Agents can discover, query, execute, edit, and what they are prohibited from touching. It supports RBAC, row filters, allow-lists, read-only sources, saved queries, audit logs, and JWT/OIDC. Agents can first query gj_catalog and gj_security, then validate filter conditions, preview generated work or CodeSQL changes, and finally execute through a controlled surface—avoiding default exposure of raw SQL or shell permissions.

Pricing and open source

The page shows v3 · Apache 2.0, indicating it is an Apache 2.0 open-source project with a self-hosted option. However, it does not provide information about a hosted version, enterprise edition, support services, pricing, or payment methods.

Pros, cons, and who it is for

Its strengths are broad coverage, native MCP support, careful consideration of Agent security boundaries, and making GraphQL compilation optimization a core feature. The drawbacks are that the captured text lacks performance benchmarks, production case studies, detailed configuration references, and commercial support information. It is suitable for platform engineering, data/API teams, and development organizations that want AI assistants to securely access internal enterprise data and workflows.

Access from China

The page does not provide information about mainland China access, mirrors, payments, or compliance, so china_access can only be marked as unknown. If network or ecosystem integration is limited, alternatives to evaluate include Hasura, PostGraphile, Supabase GraphQL, Apollo Federation, or self-built MCP tool services.