Bring Pulse to your AI agents: extract, split, schema, and tables as tools they can call.The Pulse MCP server lets AI agents in Claude, Codex, VS Code, and other MCP clients work with documents (parsing them, applying schemas, splitting by topic, and pulling tables) by calling tools, the same way they call any other MCP tool. Run it either way; the tools are the same:
- Hosted: point your client at
https://mcp.runpulse.com/mcp(Streamable HTTP). Nothing to install. - Local: run
uvx pulse-mcpon your machine (stdio). Agents can then also extract local files straight from disk.
Connect a client
Connect Codex, Claude Desktop, Claude Code, or VS Code, hosted or local.
Tools & workflows
The full tool reference and end-to-end agent examples.
What is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI agents discover and call external tools. Instead of you wiring HTTP calls into your application, the agent reads each tool’s plain-English description, decides which tool to call, fills in the arguments, and chains the results together on its own. That makes Pulse available wherever your agent already lives: ask Claude to “extract the totals from this invoice,” and it will call Pulse’sextract and apply_schema tools and
hand you back structured JSON, without you writing any extraction code.
When to use the MCP server
Use the MCP server
You’re working inside an agent or chat client (Claude, Codex, an MCP-enabled
app) and want it to read and structure documents on demand, deciding the steps itself.
Use the SDK / API
You’re building a deterministic pipeline or batch job in your own code. The
Python / TypeScript SDKs and REST API
give you precise, repeatable control.
Tools
The server exposes eight tools. Most agent work starts withextract and then chains the
returned extraction_id into the others.
| Tool | What it does |
|---|---|
extract | Parse a document at a URL (or, locally, a file path) into markdown (and optionally HTML, figures, chunks). |
apply_schema | Apply a JSON schema to a prior extraction to get structured JSON. |
generate_schema | AI-generate or refine a JSON extraction schema from a prompt. |
split_document | Split a prior extraction into topic-based page ranges. |
extract_tables | Pull tables out of a prior extraction (HTML or markdown). |
batch_extract | Extract many documents in one asynchronous batch. |
run_pipeline | Run a saved multi-step Pulse pipeline on a document. |
get_job | Poll any asynchronous job by job_id for its status and result. |
Next steps
Connect a client
Configuration for Codex, Claude Desktop, Claude Code, and VS Code.
Get an API key
Create a key in the Pulse Platform to authenticate the connector.