AI Agents

Get started building an agent that uses Andi API to understand definitions and populate workspace data

Use Andi API to let an agent ask the user what they want, inspect the workspace, and create or update data.

What is AndiBot?

AndiBot is an agent pattern on top of Andi API. It can:

ask the user what they want to do
read object definitions
create definitions if needed
read, create, and update records

First message to the user

A good first message:

Tell me what you want to do in Andi, and share your ANDI_API_KEY so I can inspect your workspace and prepare the right objects and data.

Authentication

The agent needs a workspace API key.

create it in the workspace API Keys page
send it as Authorization: Bearer <key>
store it as ANDI_API_KEY

Example:

curl "https://andiapi.com/api/objects" \
  -H "Authorization: Bearer $ANDI_API_KEY"

Flow

1. Ask what the user wants

Ask what they want to create, update, track, or load.

2. Read the object definitions

Read the workspace model first.

Object Definitions API
GET https://andiapi.com/api/objects

This tells the agent which objects exist and which fields they need.

If the object does not exist yet:

POST https://andiapi.com/api/objects

3. Populate the data

Then read or write records.

Object Data API
POST https://andiapi.com/api/objects/:objectHandle/data/upsert-many
PATCH https://andiapi.com/api/objects/:objectHandle/data/patch-many
GET https://andiapi.com/api/objects/:objectHandle/query

Minimal loop

1. Ask the user what they want to do.
2. Ask for the API key.
3. Call GET https://andiapi.com/api/objects.
4. Create a definition if the needed object does not exist yet.
5. Choose the relevant object definition.
6. Ask for any missing data required by that object.
7. Call the object data endpoint to insert or update records.
8. Confirm what was created or changed.