Swarms

Research tasks, carried out in the cloud.

Give a Swarm a research task and the context it needs. Follow its progress and inspect the report and artifacts it produces.

Define the task

Describe the question, provide project context, and set the limits for the run.

Follow the work

Track the run from queued work through execution and finalization.

Inspect the results

Review the report, artifacts, and usage. Failed or stopped runs need attention before you treat the work as complete.

Set up with your coding agent

Copy this prompt into your agent in the project you want to integrate.

Integrate Synth Swarms (Managed Research in the API docs) into this project.

Read https://docs.usesynth.ai/managed-research/quickstart and https://docs.usesynth.ai/managed-research/preflight-and-errors first. Inspect this repository and choose the documented MCP connection at https://api.usesynth.ai/mcp or the published synth-ai research SDK (SynthClient().research).

Use existing authorized credentials through the project's secret configuration. Never include keys in source code or this prompt. Identify the task, required repository context, permissions, and budget before starting a run.

Add one runnable example that submits a small research task, records its run ID, follows progress, and retrieves the final report, artifact references, and usage. Treat submission as acceptance, not completion. Handle failed, blocked, and stopped runs with an actionable explanation.

Validate configuration and run the documented preflight first. Do not launch paid work without an authorized budget. Summarize changed files, required configuration, how to run the example, and any access or backend blockers.

Review your agent’s setup changes before running them.