Most "AI for code" tools have a single chat agent. fastpace ships seven specialists, each with explicit scope and a daily job to do. They don't replace your developer — they multiply them.
/onboarder Reads your CLAUDE.md, scans the codebase, generates a tour deck, identifies first tasks. Day-1 onboarding becomes hour-1.
/architect Reads your PRD, produces entity-relationship-decisions, sizes the work as waves → epics → stories. The agent your senior engineers wish they had time to be.
/code-reviewer Knows your conventions because they live in fastpace/context/. Catches "this contradicts ADR-007" before a human reviewer has to.
/security-reviewer Cross-references SAST findings, redaction patterns, and runtime guard rules. Pairs well with audit-gate.
/evidence-curator Finds the run manifest, audit-chain entry, and provenance trailer for any AI-generated change. "Show me what produced this code" becomes one command.
/context-curator Watches for stale ADR references, missing pattern docs, terminology drift. Flags context decay before it bites the next AI session.
/debugger Walks the audit chain, finds the run manifest that produced the failing line, suggests a fix that respects the original constraints.
Bundled agents work out of the box. Add your own in .claude/agents/ with frontmatter scope. Same runtime, your conventions.
