ScalixScalix Docs
Sign up

Build with AI agents

Scalix is agent-native — operate the whole platform from any AI agent

Scalix Cloud is agent-native: every capability a human reaches through the SDK or CLI is also exposed over a machine-readable HTTP surface an AI agent can call. An agent can provision a database, run a query, deploy a function, and ship a service — autonomously, behind one token.

What makes the platform agent-first:

  • One key. A single API key authenticates every service — database, functions, storage, AI, and the rest.
  • Machine-readable errors. Every error carries a stable code, a human-readable error message, and a request_id for correlation — agents branch on codes instead of parsing prose.
  • Rate-limit headers. Responses carry X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset, so agents can pace themselves instead of retry-looping.

There are four surfaces an agent works with.

1. The REST API — the whole platform

Every platform capability is a documented REST endpoint under https://api.scalix.world. The full OpenAPI 3.1 spec is public and live at https://api.scalix.world/openapi.json — point a code generator or an agent's HTTP tooling at it and the entire surface is discoverable:

bash
curl https://api.scalix.world/openapi.json

See the API reference for every endpoint.

2. /v1/me — capability discovery

GET /v1/me describes the token making the call: identity (org and project), effective permissions and scopes, the tools those scopes unlock, and plan limits including the rate limit. It's the first call an agent should make:

bash
curl https://api.scalix.world/v1/me \
  -H "Authorization: Bearer $SCALIX_API_KEY"

3. Tools — the REST tool surface

The gateway serves a tool surface for agents: GET /api/v1/mcp returns tool definitions (name, description, JSON-Schema input) filtered to your key's scopes, and POST /api/v1/mcp/call invokes one:

bash
curl -X POST https://api.scalix.world/api/v1/mcp/call \
  -H "Authorization: Bearer $SCALIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "nl_to_sql", "arguments": {"question": "how many users signed up today?"}}'

The catalog covers 40+ tools across the whole platform — database, storage, functions, compute, KV, events, cron, domains, builds, auth, and search. MCP clients (Claude Code, Cursor, custom agents) can also connect directly to the standard MCP endpoint at https://api.scalix.world/v1/mcp (JSON-RPC 2.0). See the tool reference for the catalog, client setup, and calling conventions.

4. SDK — build agents that use Scalix

When you're writing your own agent, call Scalix directly. The chat endpoint's request/response envelope is OpenAI-compatible and serves the Scalix Lumio models, and the platform services are one typed operation away on the same client:

python
import os
from scalix_sdk import AuthenticatedClient
from scalix_sdk.generated.api.ai import chat_completion
from scalix_sdk.generated.api.database import execute_sql
from scalix_sdk.generated.models import ChatCompletionRequest, ChatMessage, SqlRequest
 
client = AuthenticatedClient(
    base_url="https://api.scalix.world",
    token=os.environ["SCALIX_API_KEY"],
)
 
# the agent reasons with a Scalix model...
reply = chat_completion.sync(
    client=client,
    body=ChatCompletionRequest(
        model="scalix-lumio-lite",
        messages=[ChatMessage(role="user", content="Summarize today's signups")],
    ),
)
 
# ...and acts on the platform with the same client
result = execute_sql.sync(
    client=client,
    body=SqlRequest(
        query="SELECT count(*) FROM users WHERE created_at > now() - interval '1 day'",
    ),
)

See the SDK docs for the TypeScript equivalent and the full service walkthrough.

Scalix's own agent

Scalix Coder is the coding agent we build on this surface — a worked example of an agent that edits code, runs commands, and deploys to Scalix Cloud.

Next: the tool reference, the API reference, or the SDK docs to build your own.