Summon a self-evolving AI team into your codebase.
Summon reads the repo first, then generates agents that fit the stack you already have.
For developers and personal AI-agent users who want
personalized, inspectable, repo-native output
instead of generic prompts.
Start from the repo, not the persona.
It detects the stack, asks only for missing context, and writes a self-evolving agent system back into the project.
- Detect the stackFrameworks, ORMs, databases, CI, and deployment clues come from the codebase itself.
- Ask only what mattersSummon fills the gaps static analysis cannot infer well, then stops.
- Self-evolving architectureAgents learn through shared evolution files, so patterns and decisions improve future work instead of disappearing.
- Write inspectable outputAgents and shared memory live as plain files in the repo.
.claude/
├── agents/
│ ├── engineering-backend-engineer.md
│ ├── quality-code-reviewer.md
│ └── orchestrator.md
└── evolution/
├── patterns.md
└── decisions.md
Install once. Run it in the repo.
Use the repo if you want the full docs. Use the command below if you already know what you want.
npx skills add secondorderai/summon-skill
summon a backend engineer for this project create a security auditor agent set up an agent team for this repo
-
What you get
.claude/agents/and.claude/evolution/files you can review and commit. - Who it coversBackend, frontend, QA, security, architecture, and custom roles from plain language.
- Compound EngineeringEach task leaves reusable knowledge behind, so every agent and every future task benefits from previous work.
- Why it stays usefulPatterns and decisions accumulate instead of resetting every run.