For developers

Build your own agent. Skip the reliability tax.

The Renderell API gives you the parts of an agent platform you don't want to write yourself: model routing, memory, typed tool use, verification, approvals, audit, rollback. Bring the workflow logic; we handle the loop.

Get an API key See the surface

Quickstart

A custom agent, top to bottom.

Define the agent in code or via the Console. Register the tools it can call. Set approval rules. Run it. The reliability layer — routing, memory, verification, rollback — is on by default.

REST for synchronous runs, WebSocket for streaming, Webhooks for long jobs.
Typed SDKs for TypeScript, Python, Go, and Java — generated from one spec.
OpenAPI 3.1 schema, Postman collection, Terraform provider.
Idempotency keys, structured errors, retries with exponential backoff.

// A custom renewal-watcher agent

import { Renderell } from "@renderell/sdk";

const r = new Renderell({ apiKey: process.env.RENDERELL_KEY });

const agent = await r.agents.create({

name: "renewal_watcher",

trigger: { type: "schedule", cron: "0 8 * * 1" },

tools: ["salesforce", "slack", "calendar"],

policy: { approval: "slack:#renewals" },

memory: "per_account",

guardrails: { discount_max_pct: 20 },

});

await r.agents.deploy(agent.id);

API surface

Everything you need to ship a custom agent

🤖

/agents

Create, deploy, and version agents. Pin a version, ship the same behavior forever; or pin to "latest" and ride the upgrades.

🧩

/tools

🧠

/memory

Per-user, per-account, per-workflow memory stores. Glossaries, preferences, prior decisions — addressable by key.

✅

/approvals

Programmatic control of the approval inbox. Read pending actions, approve, reject, or send back with comments.

📊

/runs

Every agent run is a first-class resource. Stream events, fetch traces, replay actions, roll back.

🔐

/policies

Spend limits, scope controls, model allowlists, and environment promotion. Everything CISOs ask for, exposed as code.