LlamaParse changes shipped in June 2026.
The LlamaParse MCP server now covers Extract and Index alongside Parse, Classify, and Split, with a scoped endpoint per product. This lets an agent get access to just the part of the platform it needs, and run several in parallel. See MCP.
Enterprise organizations can now tag parse and extract requests with custom labels, carried through to usage metrics. This lets billing data be sliced by team, environment, or customer. See Usage tags.
Parse jobs now accept a user_metadata object at creation, for whatever context your
application needs to keep with the job. It’s returned unchanged on every get and list
response, and can’t be changed after the job is created. See
Configuring Parse.
LlamaParse now ships official SDKs for Go and Java, plus a command-line client, alongside the existing Python and TypeScript SDKs. All are generated from the same API definition, so they share the same surface and stay in step with new releases. See the API reference.
The parse API now accepts a configuration_id, so a saved configuration can be applied
to a job directly instead of repeating every option in the request. The other APIs
already supported this; Parse now matches them. See the
parse request reference.
