cortex documentation
A terminal coding agent built on the axon runtime. Streaming chat, tool dispatch, append-only sessions, and secondary-LLM context pruning — all from your terminal.
Documentation map
cortex/
├── getting-started/
│ ├── install — go install · pre-built binaries · build from source
│ └── quick-start — env vars · interactive & non-interactive · first config
├── guides/
│ ├── configuration — the three-layer cascade · fields · known providers
│ ├── interactive — the transcript · Ctrl-C · the banner
│ ├── non-interactive — stdout/stderr split · iteration limits · piping
│ ├── commands — /new · /undo · /cd · /session · /help · /quit
│ └── keyboard — every key the terminal binds
├── concepts/
│ ├── architecture — axon owns the loop · cortex owns the product · two boundary rules
│ ├── pruner — the secondary model · what it sees · when to skip it
│ ├── trace-log — the third event sink · both models tapped · why it exists
│ └── mcp-servers — MCP protocol · auto-discovery · multiple servers
├── playground/
│ ├── overview — run.sh & trace.sh · the run directory · the six views
│ ├── running-tasks — step by step · what a run keeps · comparing runs
│ ├── reading-traces — timeline · prompt · system · tools · usage · errors
│ └── tasks — ledger · report · pipeline · what the playground found
└── reference/
├── cli — --config · --prompt · --version · exit codes
├── config-file — every field · pruner · mcp_servers · env overrides
└── api-keys — resolution order · providers.json · the leading $Start here
New to cortex? Begin with Install, thenQuick start. Once you're running, readArchitecture to understand the layering before you change anything.