Script MITHUNAI from the command line

MITHUNAI does not ship a command-line tool: no mithunai binary, npm package or Homebrew formula is published by MITHUNAI, and commands such as mithunai login or mithunai sync do not exist. Everything a CLI would do is one HTTP request away: use curl (or any HTTP client) against the HTTP API with an API key. Set two environment variables first. Every example uses them.

export MITHUNAI_URL="https://app.mithunai.com"  # your MITHUNAI host
export MITHUNAI_API_KEY="mithunai_sk_..."          # an API key; keep it in a secret store

Common tasks

Task	Request
List collections and sources	`GET /knowledge/collections`, `GET /knowledge/sources`
Upload local files	`POST /knowledge/uploads` (multipart), then `POST /knowledge/sources` with `source_type: upload`
Re-sync a source	`POST /knowledge/sources/{id}/sync`
Check an ingestion job	`GET /knowledge/jobs/{id}`
Ask an assistant a question	`POST /conversations`, then `POST /conversations/{id}/messages`

Paths are relative to $MITHUNAI_URL/mithunai/api/v1. The API reference documents every request and response, and File uploads covers the upload flow.

Ask a question from the terminal

CONVERSATION_ID=$(curl --silent --request POST "$MITHUNAI_URL/mithunai/api/v1/conversations" \
  --header "Authorization: Bearer $MITHUNAI_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{"assistant_id": "<assistant id>"}' | jq -r .id)

curl --silent --request POST "$MITHUNAI_URL/mithunai/api/v1/conversations/$CONVERSATION_ID/messages" \
  --header "Authorization: Bearer $MITHUNAI_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{"text": "How do I revoke an API key?"}' \
  | jq '{text: .answer.text, abstained: .answer.abstained, sources: [.answer.citations[].url]}'

Re-sync from CI after your docs change

MITHUNAI does not watch repositories, and a deployment’s scheduled re-sync is off unless its operator has turned it on, so trigger a sync from the pipeline that publishes your content. A GitHub Actions example, using an API key with the Editor or Knowledge operator role:

.github/workflows/mithunai-sync.yml

name: Re-sync MITHUNAI
on:
  push:
    branches: [main]
    paths: ['docs/**']

jobs:
  sync:
    runs-on: ubuntu-latest
    steps:
      - name: Sync the documentation source and wait for it
        env:
          MITHUNAI_URL: ${{ vars.MITHUNAI_URL }}
          MITHUNAI_API_KEY: ${{ secrets.MITHUNAI_API_KEY }}
          SOURCE_ID: ${{ vars.MITHUNAI_SOURCE_ID }}
        run: |
          api() { curl --silent --fail --header "Authorization: Bearer $MITHUNAI_API_KEY" "$@"; }
          JOB_ID=$(api --request POST "$MITHUNAI_URL/mithunai/api/v1/knowledge/sources/$SOURCE_ID/sync" | jq -r .id)
          for _ in $(seq 60); do
            JOB=$(api "$MITHUNAI_URL/mithunai/api/v1/knowledge/jobs/$JOB_ID")
            [ "$(echo "$JOB" | jq -r .is_active)" = "false" ] && break
            sleep 10
          done
          echo "$JOB" | jq '{status, error_code}'
          [ "$(echo "$JOB" | jq -r .status)" = "succeeded" ]

A sync is idempotent while a run is active, so a second push during a sync gets the running job back rather than starting another. See Ingestion.

For AI coding tools such as Claude Code, Cursor and VS Code, connect the MCP server instead. It gives them read-only tools to search your knowledge and ask your assistants.

MITHUNAI RAG

Knowledge sources

Answers and assistants

Deploy

Analyze

Script MITHUNAI from the command line

Common tasks

Ask a question from the terminal

Re-sync from CI after your docs change

​Common tasks

​Ask a question from the terminal

​Re-sync from CI after your docs change

Common tasks

Ask a question from the terminal

Re-sync from CI after your docs change