GraphJin MCP
GraphJin
GraphJin's MCP server gives an agent safe, schema-aware access to SQL and MongoDB instead of raw database credentials
- 3,236 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Lets the agent run queries against a working database, including writes where that is permitted
Install
- Works in
Manual install
/plugin marketplace add dosco/graphjin
/plugin install graphjin-mcp@graphjinThe plugin registers the graphjin MCP server; the graphjin binary must be installed for it to work.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Lets the agent run queries against a working database, including writes where that is permitted
This is third-party code. Review the repository files before installing.
What it does
GraphJin compiles GraphQL queries into SQL or MongoDB queries and exposes the same capability as MCP tools. Before querying anything, the agent calls query_catalog and gets the tables, relationships and examples, so it isn't guessing at the schema blind. Access is bounded by an allow-list of queries and a read-only policy where configured, rather than by handing over full database credentials. A built-in server-side agent is also available: one call to ask_graphjin_agent runs the discovery loop itself and returns a typed, evidence-backed answer.
Who it is for. For developers and data analysts who need to give an agent access to a working database without handing out raw credentials.
Good fit when
- The agent needs to answer questions from PostgreSQL, MySQL, MongoDB or another supported database
- You need an upfront catalog of tables and relationships so the agent isn't writing arbitrary SQL blind
- You need to bound the agent to an allow-list of queries and a read-only mode
Not a fit when
- You just need a plain direct SQL client without a GraphQL and MCP layer on top
- Your database isn't on GraphJin's supported list
Example request
Find every customer in the database with an overdue invoice from last quarterMCP
- Transport
- stdio
- Authentication
- not required
Environment variables
GRAPHJIN_CONFIG_PATH- Path to the GraphJin configuration folder, defaults to ./config.
Limitations
The graphjin binary must be on PATH and run with access to the project configuration (GRAPHJIN_CONFIG_PATH). GraphJin is primarily a GraphQL-to-SQL compiler and runtime for building applications, and MCP is one mode built on the same core, so wiring it up needs its own schema configuration.
How to disable. Remove the graphjin-mcp plugin via /plugin in Claude Code, or delete the graphjin entry from your agent's mcpServers configuration.
FAQ
Which databases are supported?
PostgreSQL, MySQL, MongoDB, SQLite, Oracle, MSSQL and several cloud databases; the full list is in the project README.
Does the agent need raw database passwords?
No, GraphJin holds the database connection itself, and the agent works through MCP tools and allow-listed queries.
Discussion
No questions or reviews yet.
Related
MCP serversArcadeDB ships a built-in MCP server: an agent reads graph, document and vector data, and writes are opt-in.
A knowledge graph with an MCP server: the agent searches concepts by meaning and sees how well each is supported.
An agent queries GreptimeDB metrics, logs and traces with SQL, TQL and RANGE, with write protection and data masking
The official Graphlit MCP server: an agent collects material from Slack, GitHub, mail and websites into a knowledge base and searches it
A Neo4j MCP server and skill: the agent runs Graph Data Science algorithms on a graph, from centrality to embeddings
MCP server that gives an agent access to MySQL: schema inspection and queries, read-only by default, writes behind flags
README in short
The README presents GraphJin as a compiler and runtime that gives agents one governed graph over databases, files and code through GraphQL and MCP. A dedicated graphjin-mcp plugin registers the MCP server in Claude Code via the graphjin mcp command. Besides one-off queries, it supports watch subscriptions for standing questions and a gj_artifacts store for saved queries and results. Installation is available via npm, Homebrew, Docker or prebuilt Linux and Windows packages.