Everything in your workspace, over HTTP.

UPDATED 2026-10-15 · BASE URL https://api.aitizer.com

Two doors into the same workspace: a REST API for your software and an MCP server for AI tools (Claude Code, Cursor, Codex, your own agents). Both use the same Bearer token, the same per-table permissions and the same data engine: every write runs the table's computed columns (formulas, HTTP calls, AI columns, record operations) on the server before it answers.

1. Authentication & tokens

In the app open MCP & API, create an MCP Server (it is just a named permission set), grant it tables and generate a token. Send it on every call:

Authorization: Bearer YOUR_TOKEN

Tokens can have an expiry, are shown once, and can be rotated or revoked. Every call is logged under Usage. Never put a token in a browser bundle: for browser apps use app users (SDK: db.auth) behind the Vibe relay, or your own backend.

2. Data model in 60 seconds

• A workspace is the tenant. A token belongs to exactly one workspace.
• A table has an id and a name. You can address it by either in the SDK; the REST paths use the id.
• A record is { id, fields, created_at, updated_at }. fields is keyed by column name. Select columns store and return the option id (you may write the label or the id; filters accept either), relations return the related id (or an embedded record when you ask for it), files return URLs.
• Computed columns (formula, http_request, action, json_parser, ai_agent, create/update/delete/fetch_record, loop, delay) are never written by you; the server runs them when a record is created or edited and writes the result.

3. Tables

curl -X POST https://api.aitizer.com/functions/v1/rest-api/tables \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{"name":"Cases","icon":"Scale","columns":[
        {"name":"Title","type":"text","required":true},
        {"name":"Client","type":"text"},
        {"name":"Status","type":"single_select","options":[{"id":"open","label":"Open","color":"blue"},{"id":"won","label":"Won","color":"green"}]},
        {"name":"Value","type":"number"},
        {"name":"Deadline","type":"date"}]}'

4. Columns & column types

Types. Data: text, number, date, checkbox, single_select, multi_select (options are [{id, label, color}]; colors red, orange, amber, green, teal, blue, indigo, purple, pink, gray), relation (needs relation_table_id and relation_source_column_id), file. Computed: formula (Excel-like SUM(), IF(), CONCAT(), DATEDIFF()… or JavaScript), http_request, action (a Special Field integration, 240 services), json_parser, ai_agent, create_record, update_record, delete_record, fetch_records, loop, delay. The live catalogue, including the integrations installed in your workspace, is the MCP tool check_column_types. Use the key type (alias column_type).

5. Records: read, filter, sort, embed

GET /functions/v1/rest-api/tables/{id}/records            # list (default 100, max 1000 per page)
GET /functions/v1/rest-api/tables/{id}/records/{recordId} # one record

Filters are query parameters named after the column, PostgREST style. Filtering happens in the database, so a record outside the current page is still found and count is exact.

Several parameters combine with AND. The older form filter[Field][op]=value still works and can be mixed in.

curl "https://api.aitizer.com/functions/v1/rest-api/tables/482/records?Status=eq.Open&Value=gte.1000&order=Deadline.asc&limit=20&select=Title,Value,Client(Name)" \
  -H "Authorization: Bearer $TOKEN" -H "Prefer: count=exact"

6. Records: create, update, delete, batch

Writing by filter (update or delete every record that matches) is a client-side convenience in the SDK: it resolves the ids and sends one request per record, 8 in parallel.

7. Running computed columns

Each pipeline has a time and step budget set by your plan (10 s to 10 min per row). HTTP steps go through the platform proxy with your stored credentials, so secrets never leave the server.

8. SQL (read-only)

curl -X POST https://api.aitizer.com/functions/v1/rest-api/sql \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{"query": "SELECT Title, Value FROM Cases WHERE Status = '"'"'Open'"'"' AND Value > 1000 ORDER BY Deadline LIMIT 20"}'

Grammar: SELECT (* | COUNT(*) | fields) FROM table [WHERE cond [AND cond]…] [ORDER BY f [ASC|DESC], …] [LIMIT n] [OFFSET m], with =, !=, <>, >, <, >=, <=, [NOT] LIKE, [NOT] IN (…), IS [NOT] NULL. Table and column names are the human names. It runs in a read-only transaction scoped to your workspace, with a timeout from your plan (5 to 30 s). Returns { rows, total, table, limit, offset } or { count }.

9. Schema, views & agents

Agent chat from your software: POST /functions/v1/agent-webhook?type=web&token={public_token} with { "message": "…", "session_id": "…" }; add "stream": true or Accept: text/event-stream for token-by-token SSE.

10. JavaScript SDK (aitizer.sdk)

Zero dependencies, ESM, browsers, Node 18+, Deno, Bun. Supabase-style builder on top of the REST API above. Load it from the API or install npm i aitizer.sdk.

import { createClient } from 'https://api.aitizer.com/sdk/aitizer.js'
const db = createClient('https://api.aitizer.com', process.env.AITIZER_TOKEN)

// read
const { data, error, count } = await db.from('Cases')
  .select('*, Client(Name, Email)', { count: 'exact' })
  .eq('Status', 'Open').gte('Value', 1000).ilike('Title', '%contract%')
  .order('Deadline').range(0, 19)

// write (computed columns run on the server)
await db.from('Cases').insert({ Title: 'Smith vs. Jones', Status: 'Open', Value: 12000 })
await db.from('Cases').update({ Status: 'Won' }).eq('id', 42)
await db.from('Cases').delete().eq('Status', 'Archived')     // by filter: resolves ids, 8 in parallel
const one = await db.from('Cases').select('*').eq('id', 42).single()

// SQL, schema, compute
const { data: rows } = await db.sql("SELECT Title FROM Cases WHERE Value > 1000 LIMIT 10")
await db.tables(); await db.columns('Cases'); await db.compute('Cases', 42, 'Risk score')

// app users (e-mail + password accounts stored in the app's Users table)
await db.auth.signUp({ email, password }); await db.auth.signInWithPassword({ email, password })

Rows come back flat ({ id, created_at, updated_at, ...fields }); pass { flatten: false } for the raw shape. Every call resolves to { data, error, count, status } and never throws. Types: aitizer.d.ts.

11. MCP server

Endpoint https://api.aitizer.com/functions/v1/mcp-server (Streamable HTTP, spec 2025-06-18). Claude Code, Cursor and Codex connect directly, no bridge. The live tool list is GET https://api.aitizer.com/functions/v1/mcp-server/tools; every token gets check_column_types, list_tables, list_columns, list_records, get_record, create_record, update_record, delete_record, compute_cell, list_views, semantic_search, sql_query, and the Vibe Builder credential gets the 42 management tools (tables, columns, views, schedules, agents, channels, publishing). Setup snippets for each client are in the documentation.

claude mcp add --transport http aitizer https://api.aitizer.com/functions/v1/mcp-server \
  --header "Authorization: Bearer $AITIZER_MCP_TOKEN"

12. Errors, limits & quotas

Errors are JSON: { "error": "message", "code": "CODE" }. Plan limits add "upgrade": true.

Past a monthly quota, Pro and above can pay the extra in AI credits when auto-recharge is on; otherwise the call returns 402 until the next month. Usage and limits are live in the app under Billing & usage. Full table on pricing.