Growmos MCP server
{
"mcpServers": {
"codician-growmos": {
"command": "uvx",
"args": [
"growmos"
]
}
}
}{
"mcpServers": {
"codician-growmos": {
"command": "uvx",
"args": [
"growmos"
]
}
}
}{
"mcpServers": {
"codician-growmos": {
"command": "uvx",
"args": [
"growmos"
]
}
}
}{
"servers": {
"codician-growmos": {
"command": "uvx",
"args": [
"growmos"
]
}
}
}{
"mcpServers": {
"codician-growmos": {
"command": "uvx",
"args": [
"growmos"
]
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall uvx growmosversion 0.1.5
What it does
Growmos is a knowledge graph server that builds a persistent, queryable model of your codebase by extracting facts from documentation and agent sessions, then resolving and assembling them into a graph. It lets multiple agents and sessions share a unified understanding of the codebase with full provenance tracking and fact-checking capabilities.
- Query multi-hop questions about your codebase (like "what depends on the thing we replaced in ADR-7, and who owns it?") by traversing a fact graph with citations.
- Have agents remember and link facts across sessions so development knowledge persists beyond a single context window.
- Fact-check claims made by agents against the knowledge graph to verify accuracy with edge citations.
Tools this server exposes
| Tool | Does |
|---|---|
| growmos_init | Initialize a knowledge graph in your repository, detect your agent CLI, and scan documentation. |
| growmos_next | Generate the next task packet for extraction, resolution, or review work. |
| growmos_remember | Record a fact into the graph with provenance from the current session. |
| growmos_link | Create a relation between entities in the graph with source attribution. |
| growmos_journal | Log an event or development note into the knowledge graph. |
| growmos_query | Search the graph and return answers with edge citations. |
| growmos_check | Fact-check claims against the edges in the knowledge graph. |
| growmos_eval | Measure precision, recall, and F1 of the graph against gold-set reference answers. |
| growmos_view | Open an interactive, offline explorer of the knowledge graph with force layout, search, and provenance details. |
| growmos_export | Export the knowledge graph in formats including HTML, JSON, DOT, Mermaid, Cypher, or SQL. |
| growmos_doctor | Run a 10-item checklist and health signals to verify graph integrity and quality. |
| growmos_apply | Apply a task packet result (extracted facts, resolved entities, or links) back into the graph. |
| growmos_ingest | Run the pipeline in headless mode, calling Anthropic, OpenAI, or xAI APIs for automated fact extraction and resolution. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.