OpenClaw Tools (Part 4): TTS, PDF, Lobster, and MCP

🌏 中文版

This post covers OpenClaw’s auxiliary tools: voice synthesis, document analysis, deterministic workflows, and external tool integration.

Text-to-Speech (TTS)

Converts agent responses into speech. Disabled by default.

Three Providers

Provider	Requires API Key	Description
ElevenLabs	✅ (ELEVENLABS_API_KEY)	High-quality voices
OpenAI	✅ (OPENAI_API_KEY)	OpenAI TTS API
Microsoft	❌	Uses Edge’s neural TTS, free

Auto-TTS Modes

Mode	Behavior
off (default)	Disabled
always	Convert all responses to speech
inbound	Convert only after receiving a voice message
tagged	Convert only when the response contains a [[tts]] tag

Configuration

{
  messages: {
    tts: {
      auto: "always",
      provider: "elevenlabs"
    }
  }
}

Skip Conditions

• Response already contains media
• Text is fewer than 10 characters
• Text is too long (auto-summarization before conversion can be enabled)

Slash Commands

/tts status               # Check status
/tts provider openai      # Switch provider
/tts limit 2000           # Set character limit

Settings are stored locally (per-session), not globally.

PDF Tool

Analyzes PDF documents and returns text content.

Two Modes

Mode	Behavior	Supported By
Native	Sends raw PDF bytes directly to the provider API	Anthropic, Google
Extraction fallback	Extracts text first; renders page images when text is insufficient	Other providers

Input Methods

• Local file path (supports ~ expansion)
• File URL
• HTTP/HTTPS URL (remote URLs are blocked in sandbox mode)

Parameters

Parameter	Description
pdf	Single PDF
pdfs	Multiple PDFs (up to 10)
prompt	Analysis instruction (default: Analyze this PDF document)

Limitations

• Default file size limit: 10 MB
• Extraction fallback: max 20 pages
• Native mode does not support page filtering

Lobster: Deterministic Workflow Runtime

Lobster lets OpenClaw execute multi-step tool sequences as deterministic operations.

The Problem It Solves

LLM-driven workflows have a problem: the token cost and coordination overhead of multiple tool calls is high. Lobster merges multiple tool calls into a single structured operation.

Three Core Advantages

Advantage	Description
Merged execution	One Lobster call replaces multiple tool calls
Built-in approval	Pauses before side effects, waits for human authorization
Resumable state	Paused workflows return a token that allows resumption without re-execution

Design Philosophy

Lobster uses a DSL rather than arbitrary code — deterministic + auditable. Pipelines are data, making them easy to log, diff, replay, and review.

Implementation Pattern

inbox list --json | inbox categorize --json | inbox apply --json

Chain small CLI commands with approval steps for control.

Security Mechanisms

• Enforced timeouts
• Output size limits
• Fixed executable naming
• Sandbox-aware
• Does not directly handle secrets or network calls

MCP Server Integration

OpenClaw supports MCP (Model Context Protocol) Servers to extend the agent’s toolset.

Configuration

{
  mcp: {
    servers: {
      "my-server": {
        command: "npx",
        args: ["-y", "@my-mcp/server"],
        env: { API_KEY: "..." }
      }
    }
  }
}

Management Commands

/mcp list                  # List MCP servers
/mcp status                # Check status

MCP allows OpenClaw to connect to the external tool ecosystem — databases, APIs, custom services, and more.

Media Processing

Images

• image tool: Image analysis (requires imageModel)
• image_generate tool: Image generation/editing (requires imageGenerationModel)
• Supports OpenAI, Google, fal, and other providers

Media Attachments

Inbound media is automatically copied to the sandbox workspace (media/inbound/*). Supported formats depend on the channel.

Other Tools

Tool	Description
message	Send a message to the current channel
memory_search	Semantic search over memory
memory_get	Read a specific memory file
cron	Scheduled tasks
gateway	Gateway management
nodes	Node device control
canvas	Canvas tool

Summary

OpenClaw’s toolset covers everything from voice to documents, from deterministic workflows to external MCP extensions. TTS gives the agent a “voice,” Lobster makes complex workflows predictable and auditable, and MCP opens up unlimited tool extensibility.

References

This post is compiled from the following OpenClaw source documents:

• docs/tools/tts.md — TTS voice synthesis
• docs/tools/pdf.md — PDF tool
• docs/tools/lobster.md — Lobster workflow
• docs/tools/mcp.md — MCP Server integration
• docs/tools/media.md — Media processing
• docs/tools/image.md — Image tool