How do I authenticate?

With a bearer token in the Authorization header. Keys aren't created from the app yet: ask us (or whoever runs your deployment) for one, scoped to just chat, agents, embeddings, or knowledge. Every call is attributed in the audit log.

Does chat/completions support streaming or tool calls?

Not yet. Both return a clear error today rather than being silently ignored, so you always know what you can rely on.

Which models do API calls use?

The model each agent is configured with. API calls draw on the model usage your seats include, or run on your own provider keys if you connect them.

What about agents talking to other systems?

Beam agents are also reachable over the Agent-to-Agent (A2A) protocol as JSON-RPC endpoints, and Beam itself can act as an MCP server for your other tools.

What are the default rate limits?

600 requests per minute per workspace, with a tighter lane for streaming chat. Every response carries X-RateLimit headers. If you need more, talk to us.

For developers

Your agents, callable from code.

List and message your agents, run chat completions, embed text, and search your knowledge folders, all over plain REST. Everything an agent knows in the app, it knows over the API.

Get started Talk to sales

request.sh

POST /v1/chatPOST /v1/knowledge/queryGET /v1/agents

POST /v1/chatAuthorization: Bearer ••••

The endpoints.

Small on purpose. Every route listed here is live.

GET/v1/meIntrospect the key: workspace, user, role

GET/v1/agentsList the agents your key can see, paginated

GET/v1/agents/{id}Fetch one agent's name, description, and model

POST/v1/agentsCreate an agent

POST/v1/agents/{id}/messagesSend a message to an agent, get its answer

POST/v1/chatOne-shot chat; pass agentId to pick the agent

POST/v1/chat/completionsChat completions (no streaming or tool calls yet)

POST/v1/embeddingsEmbed text with your configured embedding provider

POST/v1/knowledge/querySearch your knowledge folders

Send an Idempotency-Key header on any POST to safely retry without duplicating the request.

Built on open standards.

REST for your code, A2A for other agents, MCP for other tools.

Agents over REST

Message any workspace agent from your own software. Same instructions, knowledge, and tools as in the app.

Agent APIAccount researcherPOST /v1/agents/{id}/messages

const id = "agent-3f9c2a17b04e6d58";
const res = await fetch(
  `${BEAM_URL}/v1/agents/${id}/messages`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${BEAM_KEY}` },
  body: JSON.stringify({
    message: ""
  })
});

node agent.mjs

A2A and MCP

Every agent is an Agent-to-Agent JSON-RPC endpoint, and Beam doubles as an MCP server for the rest of your stack.

A2AMCP

Another agent sends a taskJSON-RPC 2.0

POST/api/a2a/agent-3f9c…

Questions & answers.

A small, honest REST surface over your workspace: list and message your agents, run chat completions, generate embeddings, and search your knowledge folders. Everything an agent knows (instructions, knowledge, tools) applies to API calls too.

The essential AI‑stack for your company.

Simple for business users. Ready for advanced use-cases.

Chat

Model agnostic, for everyone.

@IT Help Desk

Agents

Custom AI for recurring tasks.

Workflows

Build powerful AI automations.

Knowledge

Cited answers from your own documents.

Integrations

Integrates with the tools you use.