Use Case

Centralize and Control AI Agents

Stop scattering agent configs across code and platforms. Get a single control plane for behavior, limits, and governance — with full audit trails.

The Problem

As AI agents multiply, controlling them becomes a nightmare.

Fragmented Control

Agent configs scattered across code, env vars, and different platforms — impossible to track

No Audit Trail

When agent behavior changes, no one knows what changed, when, or why

Platform Lock-in

Different config systems for Claude, GPT, custom agents — duplicated effort everywhere

The Solution: Centralized Agent Control

One API to configure all your agents, regardless of platform.

Any platform

Single Source of Truth

One place for all agent configurations. Claude, GPT, custom agents — all controlled from OpenSink.

Complete audit trail

Full Version History

Every config change is tracked. See what changed, when, and who made the change.

< 1 second

Instant Rollback

Something broke? Roll back to any previous version with one API call.

Full transparency

Real-time Visibility

See exactly what configuration each agent is using, right now.

How It Works

1. Define Configuration Once

define-config.ts

// Define agent behavior in one place
await opensink.configurations.create({
  name: "support-agent-v2",
  
  prompts: {
    system: `You are a helpful support agent.
Be concise and solution-oriented.`,
  },
  
  tools: [
    { name: "search_docs", enabled: true },
    { name: "create_ticket", enabled: true },
    { name: "refund_order", enabled: false }, // Disabled in prod
  ],
  
  limits: {
    maxTokens: 4096,
    maxRequestsPerMinute: 30,
    maxCostPerDay: 50.00,
  },
  
  environment: {
    LOG_LEVEL: "info",
    ESCALATION_THRESHOLD: 3,
  }
});

2. Use Anywhere

use-config.ts

// Use the same config across any platform

// With Claude
const config = await opensink.configurations.get("support-agent-v2");
const response = await anthropic.messages.create({
  model: "claude-3-opus",
  max_tokens: config.limits.maxTokens,
  system: config.prompts.system,
  // ...
});

// With OpenAI
const completion = await openai.chat.completions.create({
  model: "gpt-4",
  max_tokens: config.limits.maxTokens,
  messages: [{ role: "system", content: config.prompts.system }],
  // ...
});

// Config is always in sync, regardless of platform

3. Roll Back Instantly

rollback.ts

// Something went wrong? Roll back instantly

// See version history
const versions = await opensink.configurations.versions(
  "support-agent-v2"
);
// → [{ version: 5, ... }, { version: 4, ... }, ...]

// Roll back to previous version
await opensink.configurations.rollback(
  "support-agent-v2", 
  { toVersion: 4 }
);

// Agent immediately uses the old config
// No redeploy needed

Built-in Governance

Control what your agents can do, how often, and at what cost.

Rate Limits

Control how often agents can run and how much they can spend

Tool Permissions

Enable or disable specific tools per agent, per environment

Environment Configs

Different settings for dev, staging, and production

Access Control

Control who can read and modify agent configurations

Before vs After OpenSink

Aspect	Without OpenSink	With OpenSink
Config location	Scattered in code	Centralized API
Change tracking	Git commits (if lucky)	Automatic versioning
Rollback	Redeploy entire app	One API call
Cross-platform	Duplicate configs	Single config, any agent
Runtime changes	Requires deploy	Instant updates

Take control of your AI agents

Centralized configs, full audit trails, instant rollback.

Get Started Learn about Configurations