Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/davide-desio-eleva/kirograph/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you from zero to a live kirograph_context call inside Kiro IDE. You will install the CLI, run the interactive installer to wire up Kiro, index your project, and verify that the MCP tool is available. The whole flow takes about five minutes for a medium-sized codebase.
1

Install KiroGraph

KiroGraph is not yet published to the npm registry. Install it from source with the four-command flow below.
git clone https://github.com/davide-desio-eleva/kirograph.git
cd kirograph
npm install
npm run build
sudo npm install -g .
After the global install, both kirograph and the short alias kg are available on your PATH. Run kirograph --version to confirm.
2

Run the Installer

cd into the project you want to index, then run the installer. It auto-detects Kiro and any other MCP-capable tools that are present and walks you through an interactive configuration menu.
cd your-project
kirograph install
You can also target a specific tool or skip prompts entirely:
kirograph install --target kiro    # Kiro only
kirograph install --target cursor  # Cursor only
kirograph install --all            # all detected platforms, no prompts
The installer will ask you to choose:
  • Which Kiro IDE version you are running (0.x beta or 1.x stable)
  • Whether to enable embeddings and which semantic engine to use
  • Which opt-in modules to activate (Memory, Docs, Architecture, Security, etc.)
  • Any folders to exclude from indexing (e.g. src/generated, vendor)
When it finishes, the following files are written to your project:
FilePurpose
.kiro/settings/mcp.jsonRegisters the KiroGraph MCP server with Kiro IDE and CLI
.kiro/hooks/*.json (v1.x) or .kiro/hooks/*.kiro.hook (v0.x beta)Auto-sync hooks — keeps the index fresh after each agent session
.kiro/steering/kirograph.mdTeaches Kiro how and when to use the graph tools
.kiro/agents/kirograph.jsonCustom agent config for Kiro CLI
.kirograph/config.jsonYour KiroGraph configuration (edit any time)
3

Index Your Project

If you skipped the optional index step inside the installer, run it now:
kirograph index
KiroGraph scans every file that matches your configuration’s include/exclude patterns, parses each one with tree-sitter into an AST, extracts symbols and relationships, and writes the result to a local SQLite database. You will see a live progress indicator as files are processed.
  Indexing...
  ✓ Indexed 312 files, 4,821 symbols, 11,204 edges
For incremental updates after your first full index, use sync instead — it re-processes only files whose content has changed:
kirograph sync
You do not need to run kirograph sync manually during normal Kiro sessions. The agentStop hook installed by the installer triggers an incremental sync automatically at the end of every agent session, so the graph stays current without any manual steps.
4

Make Your First Query

You can query the graph directly from the CLI before opening Kiro. Two commands are useful for a first look:
# Natural-language context for a task — same query Kiro would make
kirograph context "fix auth bug"

# Exact or fuzzy symbol lookup filtered by kind
kirograph query authenticate --kind function
kirograph context returns the symbols most relevant to the task description, their callers and callees, and the estimated impact radius. kirograph query performs a direct symbol search — useful for looking up a specific function, class, or route.
5

Open Kiro and Test

Restart Kiro IDE so it picks up the new MCP server entry in .kiro/settings/mcp.json. Once restarted, open the agent panel and try a prompt that would normally require many file reads:
What functions call authenticate, and what would break if I changed its signature?
Kiro will call kirograph_context or kirograph_impact instead of reading files one by one. You can watch the MCP tool calls in Kiro’s tool-call log to confirm the graph is being used.

​
What Happens on Index

The kirograph index command creates a .kirograph/ directory at the root of your project. Here is what lives inside it:
PathContents
.kirograph/config.jsonYour KiroGraph configuration
.kirograph/kirograph.dbThe main SQLite knowledge graph (nodes, edges, FTS index)
.kirograph/vec.dbVector store — only present when semanticEngine: "sqlite-vec"
.kirograph/orama.jsonOrama hybrid index — only when semanticEngine: "orama"
.kirograph/pglite/PGlite vector store — only when semanticEngine: "pglite"
.kirograph/lancedb/LanceDB store — only when semanticEngine: "lancedb"
.kirograph/snapshots/Architecture snapshots saved with kirograph snapshot
.kirograph/export/Browser dashboard output from kirograph export
The main database (kirograph.db) contains everything needed for structural graph traversal — symbol nodes, call/import/type edges, and full-text search indexes — with no external service required.
Add .kirograph/ to your .gitignore. The database is generated from your source and can be fully rebuilt with kirograph index at any time. Committing it adds unnecessary binary churn to your repository history.

Build docs developers (and LLMs) love

Get started for freeTalk to us