🧩 What is Airparser MCP?

What is Airparser MCP?

Airparser MCP lets AI agents work directly with your Airparser account.

Instead of reading one document in one chat, an agent can use your real Airparser setup: inboxes, schemas, parsed documents, and post-processing steps.

That means Airparser stays the extraction layer, while the agent helps with decisions, fixes, and automation.

Why this is useful

AI models can already read documents. But with Airparser MCP, they can do much more than one-off extraction.

They can work with your existing pipeline and help you improve it.

What becomes possible

  • Classify new documents and route them to the right inbox. For example, separate invoices, resumes, purchase orders, and shipping documents before they reach the right parsing workflow.

  • Import historical emails and documents from other apps. For example, an agent can pull old messages from a Gmail inbox, find attachments, and send them to Airparser automatically so you can backfill data or rebuild workflows faster.

  • Write and test post-processing Python code. An agent can assist in generating scripts to clean, normalize, enrich, and reshape parsed data.

  • Create or update extraction schemas. When a document layout changes, an agent can help adjust the schema instead of making you rebuild it by hand.

  • Combine Airparser with other agent capabilities. Once the data is structured, an agent can do more with it: send emails, fill out forms, generate PDFs, update spreadsheets, or trigger the next step in a workflow.

What Airparser MCP can do today

The current Airparser MCP server can help agents:

  • list and inspect inboxes

  • list documents in an inbox

  • upload documents and parse them

  • retrieve parsed JSON

  • generate extraction schemas

  • read and update schemas

  • read, test, save, enable, and disable post-processing steps

Who it is for

Airparser MCP is best for people who want AI agents to work with documents in a structured, repeatable way.

It is especially useful for document-heavy operations, internal tools, and AI-powered workflows.

Why not just use AI in a chat?

Because one-off prompting is not the same as working with a real parsing system.

Airparser MCP gives agents access to:

  • saved schemas

  • repeatable output structure

  • document history

  • post-processing logic

That is what makes it powerful.

How to connect

Airparser MCP is available at https://mcp.airparser.com. Step-by-step guides: Claude and ChatGPT.

Sign-in and access

  • Clients that support OAuth 2.0 (Claude, ChatGPT) open an Airparser authorization screen. Paste your Airparser API key there (find it on the Account page of the Airparser app) and click Connect to Airparser. The key is only used to authorize MCP access.

  • For clients without OAuth, send your API key in the x-api-key header.

  • The assistant sees only the inboxes and documents of the account whose API key you used.

  • Access lasts 90 days. After that, disconnect and connect again.

Available tools

  • Inboxes: list_inboxes, get_inbox, create_inbox

  • Extraction schema: get_extraction_schema, get_extraction_schema_format_guide, update_extraction_schema, update_extraction_schema_from_json_schema, generate_schema_from_document, update_fields_meta

  • Documents: upload_document_sync, list_documents, get_document

  • Post-processing (Python): get_postprocessing, get_postprocessing_runtime_rules, save_postprocessing_code, test_postprocessing_code, set_postprocessing_enabled

Tools that change your account are marked as such, so assistants can ask before running them. Replacing a schema or saved post-processing code overwrites what was saved.

New inboxes created by an assistant use the vision engine unless you ask for the text engine. See LLM Engines: Text & Vision.

Example requests

  • "Create an Airparser inbox called 'Supplier invoices', upload this invoice and show me the extracted data."

  • "Show the fields my 'Receipts' inbox extracts and add a payment_method field."

  • "List the last 10 documents in 'Supplier invoices' and tell me which ones have a total above 1000."

  • "Write a post-processing step that converts all dates to ISO format, test it on one document, and enable it if the output looks right."

Good to know

  • Files. The assistant sends the file to Airparser encoded as text, so very large files may not fit in one request. For large files, upload them in the Airparser app or by email to the inbox address.

  • Parsing time. The upload tool waits up to 15 seconds for the result. If parsing takes longer, the assistant gets the document id and can fetch the result a few seconds later; the file does not need to be uploaded twice.

  • Where documents go. A document uploaded through MCP is stored in the inbox you chose, like any other document, and follows that inbox's retention setting. In the document history it appears as "Document created (MCP: <client name>, sync)."

  • Credits. Parsing uses credits from your plan as usual. You can start free, no credit card required.

Troubleshooting

  • The assistant says Airparser is not connected, or its tools are missing: remove the connector and add it again. Access expires after 90 days.

  • The result says parsing is in progress: the document was accepted. Ask the assistant to fetch it again in a few seconds.

  • A field is wrong or empty: reprocess the document in Airparser, or ask the assistant to update the schema. The vision engine handles complex layouts better.

Still stuck? Contact support@airparser.com.

Next step

If you want to try Airparser MCP, connect it to your AI client and start with a real workflow: inspect an inbox, review parsed documents, update a schema, or test a post-processing step.

Was this article helpful?