Integration

LlamaParse integration

LlamaParse (by LlamaIndex, formerly LlamaCloud) turns documents into data for AI agents. Upload files, then parse PDFs, scans, Office files, spreadsheets, images and audio into markdown, plain text or structured page items; extract JSON that follows your own JSON Schema; classify a document against your rules; and split a concatenated PDF into sections by category. Parse, extract, classify and split wait for the job to finish and return its result. Every job spends credits billed per page ($1.25 per 1,000 credits): parse 1 (fast), 3 (cost_effective), 10 (agentic) or 45 (agentic_plus) per page, extract 8 to 60 per page by tier, classify 1 to 2, split 4; re-parsing the same file within 48 hours is free unless the cache is disabled, and retained uploaded files cost 100 credits per GB per day. Directories, batches, indexes and retrieval, chat, saved configurations, webhook configurations, the Verify alpha and the deprecated v1 parsing, pipeline and extraction-agent routes are not exposed, nor are the presigned download links of files and results. Authentication is an API key from cloud.llamaindex.ai (API Key page), sent as a Bearer token; each key belongs to one project, and every call works in that project. Accounts on the EU region (api.cloud.eu.llamaindex.ai) are not served by this integration. Rate limits are per organization, such as 5 job listings and 80 job reads per second.

Connect LlamaParse How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in LlamaParse.Built from the 32 LlamaParse tools below, on the connection you add once.

What you can do with LlamaParse

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Classify job, 4 operations

cancel_classify_jobCancel classify jobPOSTCancels a running classify job: processing stops and the job becomes CANCELLED. Jobs already COMPLETED, FAILED or CANCELLED cannot be cancelled.
delete_classify_jobDelete classify jobDELETEDeletes a classify job and its result for good. The job must be COMPLETED, FAILED or CANCELLED: stop a running one with cancel_classify_job first.
get_classify_jobGet classify jobGETReturns a classify job: status, the rules used and result (type, confidence, reasoning). Add expand=usage for the credits billed.
list_classify_jobsList classify jobsGETLists classify jobs of the project with status and result. Filter by status, job ids, configuration or date; up to 100 per page. Capped at 5 listings per second.

Extract job, 4 operations

cancel_extract_jobCancel extract jobPOSTCancels a running extract job: processing stops and the job becomes CANCELLED. Jobs already COMPLETED, FAILED or CANCELLED cannot be cancelled.
delete_extract_jobDelete extract jobDELETEDeletes an extract job and its results for good. A job still running is refused: cancel it with cancel_extract_job first.
get_extract_jobGet extract jobGETReturns an extract job: status, extract_result (the JSON), page and credit usage. Add expand=configuration or extract_metadata for the options used or citations.
list_extract_jobsList extract jobsGETLists extract jobs of the project, newest first, with status and extract_result. Filter by status, file_input, configuration_id or date. Capped at 5 listings per second.

File, 4 operations

delete_fileDelete fileDELETEDeletes an uploaded file from the project for good, which also stops its storage billing.
get_fileGet fileGETReturns one uploaded file's metadata: name, type, purpose, project, last change and expiry time.
list_filesList filesGETLists files uploaded to the project (id, name, type, purpose, expiry), 50 per page by default. Filter by file_name, file_ids or external_file_id.
upload_fileUpload filePOSTUploads a document and returns its file id (dfl-...) for parse_document, extract_data, classify_document or split_document. Retained files are billed 100 credits per GB per day.

Parse job, 4 operations

cancel_parse_jobCancel parse jobPOSTCancels a running parse job: processing stops and the job becomes CANCELLED. Jobs already COMPLETED, FAILED or CANCELLED cannot be cancelled.
delete_parse_jobDelete parse jobDELETEDeletes a parse job and its results for good. The job must be COMPLETED, FAILED or CANCELLED: stop a running one with cancel_parse_job first.
get_parse_jobGet parse jobGETReturns a parse job's status, tier, error and credits (no content). Use get_parse_markdown, get_parse_text or get_parse_items for the parsed content.
list_parse_jobsList parse jobsGETLists parse jobs of the project, newest first, with status, tier and credits used. Filter by status, job ids or creation date. Capped at 5 listings per second.

Split job, 4 operations

cancel_split_jobCancel split jobPOSTRequests cancellation of a running split job; it becomes cancelled once processing stops. Returns the job, possibly still in its current status.
delete_split_jobDelete split jobDELETEDeletes a split job and its results for good. A job not yet completed, failed or cancelled is refused (409): cancel it with cancel_split_job first.
get_split_jobGet split jobGETReturns a split job: status, categories, strategy and result.segments (category, pages, confidence).
list_split_jobsList split jobsGETLists split jobs of the project with status, categories and result segments. Filter by status, job ids or creation date.

Document, 3 operations

classify_documentClassify documentPOSTClassifies a document by your rules (result.type, confidence, reasoning). Bills 1 to 2 credits per page. Waits for the job and returns the finished result; gives up after 10 minutes.
parse_documentParse documentPOSTParses a document into markdown (job + markdown pages). Bills credits per page by tier: fast 1, cost_effective 3, agentic 10, agentic_plus 45. Waits for the job and returns the finished result; gives up after 10 minutes.
split_documentSplit documentPOSTSplits a concatenated PDF, DOCX or PPTX into page segments by category (result.segments). Bills 4 credits per page (3 if cached). Waits for the job and returns the finished result; gives up after 10 minutes.

Parse result, 3 operations

get_parse_itemsGet parse itemsGETReturns a finished parse job's structured items page by page (items.pages[].items, with bounding boxes and page size).
get_parse_markdownGet parse markdownGETReturns a finished parse job's content as markdown, page by page (markdown.pages[].markdown).
get_parse_textGet parse textGETReturns a finished parse job's content as plain text, page by page (text.pages[].text).

Extract schema, 2 operations

generate_extract_schemaGenerate extract schemaPOSTDrafts an extraction JSON Schema from a plain-language prompt and/or a sample file. Pass parameters.data_schema to extract_data as data_schema.
validate_extract_schemaValidate extract schemaPOSTChecks that a JSON Schema is accepted by extract_data and returns the schema as LlamaParse will use it. Run it before extract_data.

Project, 2 operations

get_projectGet projectGETReturns one project: id, name, organization id, default flag and dates.
list_projectsList projectsGETLists the projects of the organization (id, name, default flag). The API key works only in the project it was created in.

Data, 1 operation

extract_dataExtract dataPOSTExtracts JSON matching data_schema from a document (extract_result). Bills per page with parsing: cost_effective 8, agentic 25, agentic_plus 60. Waits for the job and returns the finished result; gives up after 10 minutes.

Parse version, 1 operation

list_parse_versionsList parse versionsGETLists the dated versions available for each parse tier and which one latest currently points to. Use a value as the version of parse_document.

Use LlamaParse from Claude Code, Cursor or Codex

Connect LlamaParse in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its LlamaParse operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect LlamaParse

Create your workspace
Sign up free. Nothing to install.
Connect LlamaParse once
Add your LlamaParse connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.