SuperAgent API

Supercharge your Agents

Put managed models, agents, memory, data, wallets, media, and workflows behind any product.

Read the docs
31managed services
52public routes
OneAgent Credit balance

Build the product. Call the rest.

Start with one service, then add more without changing the account, credit, or control model.

Agent upgrades

Upgrade the agent

Add better model choice, persistent intelligence, and durable outputs.

2 / 3
Persistent Superbrain

A Superbrain that remembers and evolves

Give every agent durable memory with fast exact and meaning-aware recall, cited answers, connected knowledge, and a reviewed history of every change.

  • Fast recall
  • Cited answers
  • Connected knowledge
  • Reviewed history

Managed product services

Run the product around it

Add managed agents, application data, wallets, and trading without building the backend yourself.

2 / 3
Agent operations

Agents that do the work

Launch research, swarms, simulations, mini apps, automations, and persistent agents from one backend.

  • Research
  • Swarms
  • Simulations
  • Mini apps
  • Automations
  • Persistent agents

Connect an agent. Give it the whole suite.

One remote connection lets an agent find and use every HivemindOS action its key permits. No custom wrapper is required.

Streamable HTTPhttps://api.hivemindos.app/mcp
List servicesSearch actionsRead safelyRun approved work

Your key still controls services, exact actions, projects, endpoint limits, approvals, and Agent Credits.

Connect your agent

Make a key that fits the job.

Choose what this app needs. The example policy updates instantly and leaves everything else out.

API key policy
{
  "label": "My product backend",
  "scopes": [
    "runs:read",
    "runs:write",
    "services:invoke",
    "services:read"
  ],
  "allowedServices": [
    "cloud-superbrain",
    "hive-research",
    "managed-agents",
    "managed-models",
    "managed-workflows",
    "miroshark",
    "swarm"
  ],
  "allowedOperations": [
    "actions.list",
    "capabilities.list",
    "services.invoke.cloud-superbrain.knowledge.graph",
    "services.invoke.cloud-superbrain.knowledge.search",
    "services.invoke.cloud-superbrain.memory.answer",
    "services.invoke.cloud-superbrain.memory.create",
    "services.invoke.cloud-superbrain.memory.evolve",
    "services.invoke.cloud-superbrain.memory.recall",
    "services.invoke.cloud-superbrain.memory.search",
    "services.invoke.hive-research.analyses.create",
    "services.invoke.managed-agents.agents.chat",
    "services.invoke.managed-agents.agents.create",
    "services.invoke.managed-models.chat.create",
    "services.invoke.managed-models.models.list",
    "services.invoke.managed-workflows.routines.run",
    "services.invoke.miroshark.runs.create",
    "services.invoke.swarm.runs.create",
    "services.list"
  ],
  "limits": {
    "*": {
      "requestsPerHour": 1000,
      "maxConcurrent": 10
    }
  }
}

This creates a server-side key policy. It does not create or expose a real key on this page.

See the advanced control model

Exact access

Allow or exclude services and individual operations. Child keys can only become narrower.

Endpoint limits

Set request, concurrency, and supported daily Agent Credit limits for exact operations.

Separate approval

Keep high-trust approval authority away from the app that proposes or executes the action.

Your app stays in control.

HivemindOS handles managed execution. Your product chooses the permissions, price boundary, and next action.

  1. Authorize the job

    A scoped key and project decide which work is allowed.

  2. Run the capability

    Call it directly or keep longer work as a durable run.

  3. Review protected actions

    Wallet moves, publishing, and other high-trust work can require a separate approval.

  4. Receive the result

    Read the response, download an artifact, or receive a signed webhook.

Build one feature. Keep the whole suite within reach.