Usage
Kaimon's terminal UI has seven tabs, each accessible by pressing its number key. This page walks through each tab.
| Key | Tab | Purpose |
|---|---|---|
1 | Server | Live server logs and status |
2 | Sessions | Connected Julia REPLs |
3 | Activity | Tool call history and results |
4 | Search | Semantic code search |
5 | Tests | Run and browse test results |
6 | Config | MCP client setup and onboarding |
7 | Advanced | Stress testing and diagnostics |
Press q to quit, ? to show a key reference overlay.
1 — Server
The Server tab shows live output from the Kaimon MCP server: incoming requests, authentication events, errors, and internal diagnostics. Use it to confirm that clients are connecting and requests are being routed correctly.
The status bar at the top shows the server address and port.
2 — Sessions
The Sessions tab lists all Julia REPLs currently connected through the Gate.
Each entry shows:
The 8-character session key (used to target tools at a specific REPL)
Connection status and heartbeat health
Julia version and active project
When multiple sessions are connected, tools that execute code require a ses or session parameter to specify the target. See Sessions for routing details.
3 — Activity
The Activity tab is a real-time feed of every tool call the MCP server has handled.
The left pane shows a scrollable list of tool call records. Each entry shows the tool name, session target, timestamp, and result status (success or error).
Select a record with ↑/↓ and press Enter to open the detail pane on the right, which shows the full input arguments, the return value, and any error message.
Key Reference
| Key | Action |
|---|---|
↑ / ↓ | Navigate entries |
Enter | Open detail pane |
Esc | Close detail pane |
Tab | Cycle pane focus |
4 — Search
The Search tab manages semantic vector indexes over your codebase and runs natural language queries against them.
The tab has three panes:
Status (top) — Qdrant and Ollama health, active collection, embedding model
Query (middle) — type your search query here
Results (bottom) — ranked code chunks with relevance scores
Press / to focus the query input, type a natural language description, and press Enter. Use Tab to move to results and ↑/↓ to scroll.
Press i to index the current project, m to open the Collection Manager, and o to configure the embedding model.
For full details — key reference, model options, collection management — see Semantic Search.
5 — Tests
The Tests tab is primarily a display surface for test runs triggered by the AI agent via the run_tests tool. When the agent runs your tests, results stream in here in real time — you can watch progress, see which testsets pass or fail, and read failure details without leaving the TUI.
You can also trigger a run manually with r. If multiple gate sessions with a test/runtests.jl are connected, a picker dialog appears so you can choose which project to test.
Results stream in as they complete. When all tests finish, the pane shows a pass/fail summary tree.
Select a failed test with ↑/↓ and press Space to expand it. The right pane shows the failure message and backtrace. Toggle between the structured tree view and raw output with o.
Key Reference
| Key | Action |
|---|---|
r | Run tests (opens session picker if needed) |
o | Toggle tree / raw output view |
F | Toggle follow mode (auto-select newest run) |
x | Cancel a running test |
↑ / ↓ | Navigate runs or tree nodes |
Space | Expand / collapse tree node |
Esc | Cancel running test |
Tab | Cycle pane focus |
6 — Config
The Config tab handles all MCP client setup. It generates and writes configuration files for supported editors and CLI tools.
Supported clients: Claude Code, VS Code (Copilot / Continue), Cursor, Gemini CLI, KiloCode.
Press i to open the client list and select a target. Press Enter to write the config. Kaimon writes the correct format for each client including your API key if running in authenticated mode.
The onboarding flow (o) creates a .julia-startup.jl file for per-project auto-connect, and the global gate option (g) appends the same snippet to ~/.julia/config/startup.jl so every Julia session connects automatically.
See Getting Started for the full onboarding walkthrough.
7 — Advanced
The Advanced tab provides stress testing and load diagnostics. It lets you fire configurable bursts of concurrent tool calls against the running server to measure throughput and latency under load.
Use this tab to verify that the server handles parallel requests correctly when multiple agents or sessions are active simultaneously.