llms.txt

# primkit

> Infrastructure primitives for AI agents — tasks, state, knowledge graphs, and work queues. Single Go binaries with embedded SQLite. CLI, HTTP API, and MCP interfaces. Zero dependencies, zero config.

primkit provides four primitives that agents use directly via CLI or through MCP integration:

- **taskprim** — task management with lifecycle tracking, lists, labels, per-agent seen-tracking, and structural task-to-task dependencies (cycle detection, frontier queries)
- **stateprim** — key-value state, dedup lookups, and immutable append logs in namespaced storage
- **knowledgeprim** — knowledge graphs with typed entities, weighted contextualized edges, hybrid search (FTS5 + vector + RRF), and discovery operations
- **queueprim** — persistent work queues with priority, retries, dead-letter support, and atomic dequeue with visibility timeouts

Each primitive is a single binary that auto-creates its SQLite database on first use. No server required for CLI usage.

## Documentation

- [README](https://github.com/propifly/primkit/blob/main/README.md): Installation, quick start, API endpoints, MCP configuration
- [Agent Reference](https://github.com/propifly/primkit/blob/main/docs/agent-reference.md): Structured command tables, JSON schemas, decision trees, error patterns for all four primitives
- [AGENTS.md](https://github.com/propifly/primkit/blob/main/AGENTS.md): Instructions for AI agents developing or using primkit
- [knowledgeprim Guide](https://github.com/propifly/primkit/blob/main/docs/knowledgeprim.md): Entity types, relationships, edge context, search strategy, discovery workflows
- [queueprim Guide](https://github.com/propifly/primkit/blob/main/docs/queueprim.md): Visibility timeout, worker loop, priority and queue design, retry and dead-letter strategy
- [Configuration](https://github.com/propifly/primkit/blob/main/docs/configuration.md): Full YAML spec, environment variable overrides, embedding provider setup
- [Architecture](https://github.com/propifly/primkit/blob/main/docs/architecture.md): Layered design, store interfaces, data model, replication

## Optional

- [Contributing](https://github.com/propifly/primkit/blob/main/CONTRIBUTING.md): Development setup, code style, PR process
- [Changelog](https://github.com/propifly/primkit/blob/main/CHANGELOG.md): Release history
- [Setup Guide](https://github.com/propifly/primkit/blob/main/SETUP.md): R2/S3 replication setup, MCP configuration walkthroughs