Installation
Install DevGraph globally or run it directly withbunx:
1. Add service blocks where they help
devgraph build scans your repo by default. Blocks are optional, but they are the authoritative service layer when you want named services, commands, APIs, environment variables, or explicit owned paths.
Example:
2. Build the graph
3. Check outputs
Outputs are generated in.devgraph/:
| File | Description |
|---|---|
graph.json | Machine-readable output containing top-level services, apis, and knowledgeGraph |
summary.md | Human-readable service overview |
GRAPH_REPORT.md | Hybrid graph analysis with god nodes, surprising connections, bridge nodes, and coverage gaps |
agents/*.md | Per-service context files for coding assistants |
system.mmd | Mermaid diagram of the build pipeline and service relationships |
codemap.mmd | Mermaid diagram of owned paths and file relationships |
4. Query the graph
Usedevgraph query when you want a focused answer from the built graph:
5. Visualize with Studio
Launch the interactive graph viewer:- service and file nodes
- search plus graph filters
- community, provenance, and ownership views
- node detail panels with evidence and neighbors
6. Generate agent skills
Generate reusable skills for AI coding tools:.skills/ with architecture, service, and orchestration skills. The orchestration skill tells agents how to use GRAPH_REPORT.md, run focused devgraph query prompts, and produce a Markdown handoff brief for another agent.
Next Steps
Build Command
Learn about
.devgraphignore, --compare, and hybrid outputs.Query Command
Use graph retrieval instead of broad file search.
Agent Skills
Generate architecture, service, and orchestration context for AI agents.