Ask your data in plain English.

Most teams have the data but not the analyst time. Dashboards answer the questions you planned for; everything else becomes a ticket. Querion turns the long tail of "can someone just pull..." into a chat message, while staying safe enough to point at production.

If a platform has an API, Querion can read it. You do not write code. You give Querion three things, and it pulls the data and structures it into the answer for you.

• Add the key and secret to the environment (a .env file or your secret manager). Querion never stores secrets in config.
• Describe the API in a short markdown file: the read endpoints and what they return, the way you would brief a new analyst.
• Connect your Postgres with a read-only role. That is it. Querion reads your schema and your endpoints and starts answering.

sources:
  # Stripe, your billing, your 3PL, your CRM... any read API
  - name: stripe
    base_url: https://api.stripe.com
    auth_header: "Authorization: Bearer ${STRIPE_KEY}"
    safe_get:                # only these read paths are allowed
      - ^/v1/charges
      - ^/v1/customers
      - ^/v1/invoices
    docs: docs/sources/stripe.md

Now a single question can span Postgres and Stripe together. Querion fetches from each source, reconciles them, and returns one structured answer with the table, the chart, and the query behind it. Repeat the block for every platform you run.

Querion lives where your team already talks. A thin bridge forwards a message to Querion and posts the answer back, so the same analyst works in any instant-messaging app: Slack, WhatsApp, Microsoft Teams, Discord, or your own.

The bridge is small: it receives a message, calls Querion, and replies. Daily per-user limits, a write-request firewall, and read-only sources travel with Querion, so opening it to a channel stays safe.

From zero to asking questions in a few minutes. You need Python 3.9+, a Postgres database, and the Claude Code CLI logged in once on the host.

git clone https://github.com/anishfyi/querion.git
cd querion
pip install -e ".[all]"

claude   # log in once (browser / your plan)

cp querion.example.yaml querion.yaml
cp .env.example .env
# edit .env with your read-only Postgres DSN + API keys

CREATE ROLE querion_ro LOGIN PASSWORD 'choose-a-strong-one';
GRANT CONNECT ON DATABASE yourdb TO querion_ro;
GRANT USAGE ON SCHEMA public TO querion_ro;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO querion_ro;
ALTER ROLE querion_ro SET default_transaction_read_only = on;

querion check                          # config + DB + CLI
querion ask "orders and GMV last 7 days"
querion serve                          # web UI on :8000

Full configuration reference, the API-source format, and the architecture are in the repo. Start from the README.

"It never writes" is true by design, with defense in depth. The only thing Querion ever produces is the answer.

Querion is far sharper when it knows your domain: what "active customer" means, which table is the source of truth, how your metrics are defined. Pair it with Trove, a companion that builds and maintains a file-based semantic layer as you work and reloads it every session. Enable it in the config and Querion folds that layer into its knowledge, so every answer speaks your business language.