@buildlogai/mcp

MCP server for Buildlog - search, follow, and record AI coding workflows.

This server gives AI agents direct access to the buildlog ecosystem, enabling them to:

• Search for relevant workflows before starting a task
• Follow proven buildlog recipes in their current context
• Record their own sessions as they work (auto-record by default!)
• Upload completed buildlogs for others to learn from

Installation

npm install -g @buildlogai/mcp

Configuration

Set your API key for uploads (optional for search/follow, required for upload):

export BUILDLOG_API_KEY=your-api-key

Auto-Recording

By default, the MCP server starts recording automatically when loaded. This ensures no workflow is lost.

To disable auto-recording:

export BUILDLOG_AUTO_RECORD=false

When a session ends with meaningful content (3+ steps), the server will suggest uploading.

Use buildlog_auto_status to check if auto-recording is active.

Proactive Suggestions

Before starting a complex task, use buildlog_suggest:

Use buildlog_suggest with taskDescription "Add Stripe subscription checkout to Next.js app"

The server will search for relevant workflows and recommend the best matches.

Source Attribution

All buildlogs created via MCP include source metadata:

• Tool name and version (@buildlogai/mcp)
• Client application (if detectable via MCP_CLIENT env var)

This helps track which tools contribute to the collective knowledge base.

Usage

Claude Desktop

Add to your Claude Desktop config (~/.claude/claude_desktop_config.json):

{
  "mcpServers": {
    "buildlog": {
      "command": "buildlog-mcp"
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "mcpServers": {
    "buildlog": {
      "command": "npx",
      "args": ["@buildlogai/mcp"]
    }
  }
}

OpenClaw

OpenClaw supports MCP servers natively:

{
  "mcp": {
    "servers": ["buildlog-mcp"]
  }
}

Available Tools

| Tool | Description | |------|-------------| | buildlog_search | Search buildlog.ai for relevant workflows | | buildlog_suggest | Get proactive workflow suggestions before starting a task | | buildlog_get | Fetch a specific buildlog by slug or URL | | buildlog_get_steps | Get just the steps from a buildlog | | buildlog_follow | Get prompts formatted for execution | | buildlog_record_start | Begin recording a session (or convert auto-session) | | buildlog_record_step | Log a step (prompt, action, terminal, note) | | buildlog_record_stop | End recording, get the buildlog | | buildlog_auto_status | Check auto-recording status and stats | | buildlog_upload | Upload to buildlog.ai | | buildlog_fork | Start recording from an existing buildlog |

Tool Details

buildlog_search

Search buildlog.ai for relevant workflows.

Parameters:
- query (required): Search query (e.g., "stripe nextjs integration")
- language (optional): Filter by language (e.g., "typescript")
- framework (optional): Filter by framework (e.g., "nextjs")
- limit (optional): Max results (default: 10)

buildlog_suggest

Get proactive workflow suggestions before starting a task. Call this before starting complex work.

Parameters:
- taskDescription (required): Description of what you're about to do
- contextFiles (optional): Files relevant to the task (helps improve matching)

Returns:
- Ranked list of relevant workflows
- Top recommendation with follow command
- Extracted keywords that were matched

buildlog_get

Fetch a complete buildlog by slug or URL.

Parameters:
- slug (required): Buildlog slug or full URL

buildlog_get_steps

Get just the steps from a buildlog (lighter response).

Parameters:
- slug (required): Buildlog slug or full URL

buildlog_follow

Get prompts from a buildlog formatted for execution. This is the key tool for replicating workflows.

Parameters:
- slug (required): Buildlog slug or full URL
- step (optional): Start from a specific step number

buildlog_record_start

Begin recording a new session. If auto-recording is active, this converts the auto-session to a named session.

Parameters:
- title (required): Title for the recording
- description (optional): Longer description

buildlog_record_step

Log a step to the active recording.

Parameters:
- type (required): "prompt" | "action" | "terminal" | "note"
- content (required): The step content
- metadata (optional): Additional data (filesCreated, filesModified, etc.)

buildlog_record_stop

End recording and return the buildlog.

Parameters:
- outcome (optional): "success" | "partial" | "failure"
- summary (optional): Summary of what was accomplished

buildlog_auto_status

Check if auto-recording is active and get current session stats.

Parameters: none

Returns:
- autoRecordEnabled: Whether auto-record is on
- isRecording: Whether currently recording
- isAutoSession: Whether this is an auto-started session
- stats: Step count, duration, prompt/action counts

buildlog_upload

Upload a buildlog to buildlog.ai.

Parameters:
- buildlog (required): The buildlog object
- public (optional): Whether publicly visible (default: true)

buildlog_fork

Start a new recording based on an existing buildlog.

Parameters:
- slug (required): Source buildlog slug
- title (required): Title for the new buildlog
- fromStep (optional): Only copy steps up to this number

Example Workflows

Search and Follow

Agent: Use buildlog_search to find "stripe nextjs integration"
→ Returns list of relevant buildlogs

Agent: Use buildlog_follow on buildlog "abc123"
→ Returns prompts to execute in order

Agent: Execute each prompt, adapting to current project

Record a Session

Agent: Use buildlog_record_start with title "Building auth system"

Agent: ... does work ...
Agent: Use buildlog_record_step for each significant action

Agent: Use buildlog_record_stop when done
→ Returns completed buildlog

Agent: Use buildlog_upload to share
→ Returns URL to view buildlog

Fork and Extend

Agent: Use buildlog_fork on "abc123" with title "Extended auth with MFA"
→ Starts recording with inherited steps

Agent: Use buildlog_record_step for new work
Agent: Use buildlog_record_stop
Agent: Use buildlog_upload

Environment Variables

| Variable | Description | |----------|-------------| | BUILDLOG_API_KEY | API key for uploads (get from buildlog.ai/settings) | | BUILDLOG_API_URL | Custom API URL (default: https://buildlog.ai/api) | | BUILDLOG_AUTO_RECORD | Set to "false" to disable auto-recording (default: true) | | MCP_CLIENT | Client name for source attribution (e.g., "Claude Desktop", "Cursor") |

VS Code Extension Interop

When recording, steps are also written to ~/.buildlog/agent-feed.jsonl (if the directory exists). This allows the VS Code extension to incorporate MCP-based recordings into its capture flow.

Links

• buildlog.ai
• Documentation
• GitHub
• VS Code Extension

License

MIT