Concepts
The what and why of Codebolt — the mental models you need to understand the rest of the docs.
This section is explanation, not instruction. Looking for how to do something? Try Guides. Looking for what a flag does? Try Reference.
Where to start
- Architecture — the 5-plane model and server subsystems
- Agents — what an agent actually is, and the four creation levels
- Tools and MCP — how agents act on the world
- Capabilities — packaged behaviour units
- Hooks and processors — interception and transformation
- Context and memory — what an agent "knows" at any moment
- Shadow git and checkpoints — safe rollback
- Event log — the source of truth
- Guardrails — keeping individual runs in bounds
- Multi-agent patterns — when one agent isn't enough
- Orchestration flows — how patterns become runnable graphs
- Dispute resolution — the four-layer conflict management framework
- Evals and optimization — measuring and improving agent quality over time
- The planning system — how Roadmap, Specs, UI Flow, Requirement Plan, Action Plan, and Tasks form a coherent stack
How concepts relate to the rest of the docs
| You want to… | Go to |
|---|---|
| Understand what something is | Concepts (here) |
| Use it in the desktop app or CLI | Using Codebolt |
| Follow steps to accomplish a task | Guides |
| Build custom agents/tools/providers | Build on Codebolt |
| Look up an API or schema | Reference |