MCP Integration
Works with the agent you already use.
Oynix runs as an MCP server. Any agent that speaks MCP connects in one command, then answers with the memory of your whole team. Your agent still does the reasoning. Oynix returns the facts, so retrieval costs zero tokens.
Pick yours. Get the exact command.
18 hosts supported today. Choose one for the command that wires it up and the file the config lands in.
$ oynix mcp setup --client claudeConfig lands in.mcp.json (project root), or run claude mcp add
These set up in one command. Because MCP is an open standard, any other MCP-capable assistant works too, with a config you paste once. All names and logos belong to their respective owners.
What your agent can now read.
Oynix indexes the code and the documents around it, then links the two. Your agent stops guessing from a filename and starts answering from the whole record.
Languages
Parsed into one graph, across 110 file extensions.
Document types
Turned into text and woven in beside the code they describe.
Five tools. Zero tokens.
Once connected, your agent gets purpose-built retrieval tools. Each one reads straight from the graph. There is no model on our side to bill you for, so asking is free.
ownershipWho owns this file, and who last changed it.
sourceHow the code actually works, from the graph.
contextWhy this exists: the tickets and threads behind it.
impactWhat calls this, and what breaks if you change it.
blast_radiusThe full reach of a change, across every repo.