Bootstrap
For existing codebases, trurlic bootstrap populates the decision graph autonomously. A coding agent reads the source code and records components, decisions, and patterns — no interactive dialogue needed.
Quick start
Section titled “Quick start”trurlic inittrurlic bootstrap # shows status and agent instructionstrurlic serve # start MCP server# In your coding agent:# advance(component="project", task_type="bootstrap")# follow each step until ready: truetrurlic map # review the resultHow it works
Section titled “How it works”Bootstrap is a two-phase process:
Phase 1: trurlic bootstrap — prints the current status (which components have been bootstrapped, which haven’t) and the agent instructions for running the bootstrap workflow.
Phase 2: Agent runs the bootstrap workflow — the coding agent calls advance(component="project", task_type="bootstrap") and follows the returned steps. The bootstrap task type uses an autonomous step sequence: scan_project → extract_decisions → project_rules → pattern_detection → ready.
The agent reads source code, identifies components and architectural decisions, records them via record_decision and add_component, and synthesizes patterns. All decisions are attributed as agent — flagged for human review.
Bootstrap a single component
Section titled “Bootstrap a single component”trurlic bootstrap authThis scopes the bootstrap to a single component, useful for incremental extraction.
Direct bootstrap (no agent)
Section titled “Direct bootstrap (no agent)”Use -p to run the bootstrap directly via the LLM API instead of printing agent instructions:
trurlic bootstrap -p anthropictrurlic bootstrap auth -p openai -m gpt-4oThis is a convenience mode — it runs the same workflow but uses the session module’s LLM client directly instead of relying on the coding agent’s MCP calls.
Reviewing the result
Section titled “Reviewing the result”All bootstrap-extracted decisions carry attribution = "agent" and are flagged with ⚠ in context briefs and the interactive map. Review them with:
trurlic map # visual review in browsertrurlic status # counts and health summarytrurlic check # validate graph integrityConvert agent-attributed decisions to human-attributed by revisiting them:
trurlic design auth --revisitThe revisit flow walks through existing decisions, letting you confirm, modify, or reject each one.