Skip to main content
EVOKORE// SKILLS / CLI_INTEGRATION
EVOKORE-MCPmarkdownsha 7D5B8736AAFC

CLI_INTEGRATION

documentation

EVOKORE CLI Integration

This document explains how EVOKORE integrates with supported AI CLIs beyond the raw MCP registration step.

What this covers

  • Operator-facing CLI integration patterns EVOKORE supports
  • Status line rendering and where it gets its data
  • Per-CLI integration paths (Claude Code, Gemini CLI, Cursor, Copilot/Codex)
  • Cross-CLI config sync via scripts/sync-configs.js
  • Recommended operator flow

What EVOKORE adds to CLIs

EVOKORE supports three operator-facing CLI integration patterns:

  1. MCP server registration through dist/index.js
  2. optional continuity-backed status line rendering through scripts/status.js
  3. cross-CLI registration sync through scripts/sync-configs.js

Status line integration

The EVOKORE status line is continuity-first. It summarizes:

  • branch and worktree pressure
  • session purpose
  • task pressure
  • continuity health
  • context usage when the client exposes it

Data comes from:

  • ~/.evokore/sessions/{sessionId}.json
  • current git state
  • managed Claude memory fallback when no repo-scoped live manifest exists

Claude Code

Claude Code supports a statusLine command block in ~/.claude/settings.json.

{
  "statusLine": {
    "type": "command",
    "command": "node /absolute/path/to/EVOKORE-MCP/scripts/status.js"
  }
}

Notes:

  • this repo treats the Claude Code status line as an active supported path
  • scripts/status.js is backed by scripts/status-runtime.js
  • if you also use hook-based voice forwarding, that remains separate from the status-line command

Gemini CLI

Gemini CLI can invoke the EVOKORE status line through its hook system.

{
  "enableHooks": true,
  "hooks": {
    "AfterModel": [
      {
        "type": "command",
        "command": "node /absolute/path/to/EVOKORE-MCP/scripts/status.js"
      }
    ]
  }
}

Cursor

Cursor primarily uses the MCP registration path rather than a dedicated EVOKORE status-line integration. Use the normal MCP registration flow described in SETUP.md.

Copilot CLI / Codex CLI

These CLIs now support automated MCP registration through scripts/sync-configs.js. They still do not have a native EVOKORE JSON status-line integration in this repo, so if you want a post-command status line, wrap the CLI call in your own shell function or alias.

PowerShell example:

function codex-evokore {
    codex $args
    node "/absolute/path/to/EVOKORE-MCP/scripts/status.js"
}

Cross-CLI config sync

The sync helper registers the evokore-mcp MCP entry across supported CLIs:

npm run sync:dry
npm run sync

Supported automated targets:

  • claude-code
  • claude-desktop
  • cursor
  • copilot
  • codex

gemini remains a manual command because its CLI manages MCP config through its own command surface.

Key behavior:

  • dry run is the default safety mode
  • only the evokore-mcp entry is added or updated
  • Claude Code prefers the native user MCP config at ~/.claude.json when present and falls back to ~/.claude/settings.json for older setups
  • the script resolves the canonical repo root so disposable worktrees do not leak temp dist/index.js paths into user config files
  • use --force only when you intentionally want to overwrite an existing EVOKORE entry

If you need to override the detected repo root:

EVOKORE_SYNC_PROJECT_ROOT=/absolute/path/to/EVOKORE-MCP node scripts/sync-configs.js --apply

Recommended operator flow

  1. npm run build
  2. npm run sync:dry
  3. npm run sync
  4. configure the optional status line for the CLI you use
  5. run npm run repo:audit before starting a new multi-slice repo session

See also

Last verified: 2026-05-20