Skip to main content

​
Overview

The Agents API lets you provision managed agents — configurable AI workers with their own tools, memory, and runtime behavior — and drive them programmatically. Agents run in one of two modes:

Assist Mode

Turn-based, conversational. You send a message to a session and stream back the agent’s reply — ideal for chat-style copilots embedded in your product.

Autonomous Mode

Task-based. You submit a task (“run”), the agent works independently (using tools, memory, and multi-step reasoning) and produces deliverables/artifacts you poll or subscribe to.

​
Core Concepts

The configuration object: name, system prompt, model, tool allowlist, memory settings, and compute limits (max iterations, wall-clock timeout). Created once, reused across many sessions and runs.
A conversation/work context scoped to one agent. Sessions hold message history and metadata (e.g. which external platform or end user they belong to). Assist-mode messages and autonomous-mode runs both happen inside a session.
A single autonomous task execution inside a session — e.g. “research X and produce a summary document.” Runs progress through pending → running → completed/failed and emit events and artifacts as they work.
When an agent needs to call a tool it doesn’t have native access to, it emits a pending tool call. Your integration executes the tool and reports the result back via the tool-results endpoint.

​
Base URL

https://suite.sundaypyjamas.com/api/v1/managed-agents

​
Authentication

All requests require a valid API key in the Authorization header: 
Authorization: Bearer spj_ai_your_api_key_here
Learn more about API key generation and management.

​
Resources

Agents

Create, list, update, and delete agent configurations.

Runs

Submit autonomous tasks and track their progress.

Sessions

Manage the conversational/work context an agent operates in.

Messaging

Stream assist-mode replies, read session timelines, submit tool results, and cancel runs.