Complete documentation for all MCP tools provided by gemini-kit.
Create a git checkpoint before making changes. Allows for easy rollback.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
name |
string | Yes | Descriptive name for the checkpoint |
includeUntracked |
boolean | No | Include untracked files (default: false) |
Returns:
checkpointId: Unique identifier for rollbackbranch: Name of the backup branch createdcommitHash: Git commit hash
Example:
{
"name": "before-refactor-auth",
"includeUntracked": true
}Restore the codebase to a previous checkpoint state.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
checkpointId |
string | Yes | ID from kit_create_checkpoint |
Returns:
- Success message with restored state details
Example:
{
"checkpointId": "chk-1702656000000"
}List all available checkpoints for the current project.
Parameters: None
Returns:
- Array of checkpoints with ID, name, timestamp, and commit info
Gather comprehensive project context including structure, dependencies, and recent changes.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
depth |
number | No | Directory depth to scan (default: 2) |
Returns:
structure: File/directory treepackage: Package info (if package.json exists)recentCommits: Last 5 git commits
Transfer context from one agent to another in a workflow.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
fromAgent |
string | Yes | Current agent name |
toAgent |
string | Yes | Target agent name |
context |
string | Yes | Context to pass |
artifacts |
string[] | No | File paths of related artifacts |
Example:
{
"fromAgent": "planner",
"toAgent": "coder",
"context": "Plan complete. Implement auth module with JWT.",
"artifacts": [".gemini-kit/artifacts/plan/auth-plan.md"]
}Save an artifact (plan, report, log) from agent work.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
name |
string | Yes | Artifact name |
type |
enum | Yes | plan, report, log, or other |
content |
string | Yes | Artifact content (markdown) |
Returns:
- File path where artifact was saved
Save a learning from user feedback to improve future responses.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
category |
enum | Yes | code_style, bug, preference, pattern, other |
lesson |
string | Yes | The lesson learned |
context |
string | No | Additional context |
Example:
{
"category": "code_style",
"lesson": "User prefers arrow functions over regular functions",
"context": "Applies to all JavaScript/TypeScript files"
}Retrieve learnings with optional semantic search.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
query |
string | No | Search query for relevant learnings |
category |
string | No | Filter by category |
limit |
number | No | Max results (default: 5) |
Example:
{
"query": "typescript arrow functions",
"limit": 3
}Store a proposed diff for later application (Dry Run mode).
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
file |
string | Yes | File path |
originalContent |
string | Yes | Original file content |
newContent |
string | Yes | Proposed new content |
description |
string | No | Change description |
Returns:
diffId: ID for later application- Unified diff preview
Apply a previously stored diff with conflict detection.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
diffId |
string | Yes | Diff ID from kit_store_diff |
force |
boolean | No | Force apply despite conflicts (DANGEROUS) |
Returns:
- Success/failure with conflict details if any
Index the codebase for keyword-based search.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
extensions |
string[] | No | File extensions (default: ['.ts', '.js', '.py', ...]) |
maxFiles |
number | No | Maximum files to index (default: 100) |
Returns:
- Number of files indexed
- Top files by size
Search indexed codebase by keywords, function names, or file names.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
query |
string | Yes | Search query |
limit |
number | No | Max results (default: 10) |
Returns:
- Ranked search results with file, functions, classes, and score
Start a new team session with a goal.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
goal |
string | Yes | The goal/task for the session |
sessionName |
string | No | Optional session name |
Returns:
- Session details
- Suggested workflow based on goal
- List of all available workflows
Get current team session status and progress.
Parameters: None
Returns:
- Session summary
- Completed steps
- Current agent
End the current team session.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
status |
enum | No | completed or failed (default: completed) |
Returns:
- Final session summary
Execute a predefined workflow.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
workflow |
string | Yes | Workflow name (see below) |
task |
string | Yes | Task description |
Available Workflows:
| Name | Description |
|---|---|
cook |
Full cycle: Plan → Scout → Code → Test → Review |
quickfix |
Bug fix: Debug → Code → Test |
feature |
New feature: Design → Plan → Code → Test → Docs |
refactor |
Refactoring: Scout → Plan → Code → Test → Review |
review |
Code review: Scout → Review → Security |
tdd |
Test-driven: Plan → Test (fail) → Code → Test (pass) |
docs |
Documentation: Scout → Analyze → Write → Review |
Analyze task and auto-select the best workflow.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
task |
string | Yes | Task description to analyze |
Returns:
- Recommended workflow
- Confidence score
- Reasoning
- Alternative workflows
List all available workflows.
Parameters: None
Returns:
- Array of workflows with name and description
Get history of past team sessions.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
limit |
number | No | Max sessions (default: 10) |
Returns:
- Array of past sessions with goal, status, and timing
Triggered when a new session begins. Initializes gemini-kit state.
Injects relevant learnings and project context before each agent response.
Security validation:
- Secret detection (API keys, tokens, passwords)
- Path traversal prevention
- Dangerous command blocking
Auto-runs tests after file changes to catch regressions early.
Cleanup and saves session logs.
| Code | Description |
|---|---|
| 0 | Success |
| 1 | General error |
| 2 | Security block (secret/dangerous command detected) |
Create .gemini/settings.json in your project to customize:
{
"fileExtensions": [".ts", ".js", ".tsx", ".jsx", ".py"],
"excludeDirs": ["node_modules", "dist", ".git"]
}