Skip to content

DEVELOPERS & AGENTS

Find it. Research it. Build on it.

Give your software and agents the full Specstream capability: precise requests, qualified results, researched facts, evidence, maps, and continuing work. Use the API or connect through MCP.

Recorded in the Specstream frontend · Example data

Build on the whole capability.

Read the documentation

Work with explicit intent.

Create and retain specs, prepare a scope, and inspect the interpretation. Keep qualification conditions separate from the facts to research.

Follow the work.

Use progress and events to understand what is happening. Keep partial output, completion, zero matches, and source failure distinct.

Use the result.

Retrieve retained deliveries, facts, and provenance. Export the work or continue it with more matches, more research, or a refresh.

Give agents a clear remit.

Connect through MCP

A key identifies the caller.

Manage API keys and OAuth access. A credential is not a substitute for explicit execution authority.

An authority defines the work.

Choose allowed operations, project restrictions, per-run and per-period limits, and expiration. Revoke it when the remit ends.

The same balance applies.

People, agents, and recurring work share the capacity system. A quote and its authorized limits remain the basis for paid execution.

A few useful answers.

Is the API a smaller version of the product?

The same core concepts support the website, software, and agents: specs, work scopes, evidence, deliveries, retained work, capacity, and authority.

Can an agent continue work started by a person?

Yes. Retained context gives the next operation a stable starting point, subject to the caller’s access and authority.

Can I receive events?

Customer webhooks connect work and delivery events to your systems. Manage endpoints, scope, and revocation as part of developer access.

What would be useful if you already had it?