Agents can now execute multiple independent tool calls concurrently within a single reasoning step. Instead of reading files one by one, Queue fans out tool calls in parallel and merges the results before the next model call. On large codebases this can cut run time by 40–60% for read-heavy tasks.

You can now configure a Queue agent to trigger automatically on every pull request in connected GitHub repositories. The agent reads the full diff, cross-references your codebase, and posts structured inline comments — surfacing bugs, style violations, and missing test coverage without you lifting a finger. Enable it under Settings → Integrations → GitHub.

Queue’s built-in context compressor has been overhauled. The new rolling summary strategy is 3× more aggressive at trimming verbose tool outputs while preserving the facts that matter most. Agents working across large repos now stay coherent over longer runs without hitting token limits. Configure aggressiveness in queue.yaml under context.strategy.

Queue agents can now read pages and databases, create new content, and update properties in Notion — all from a single agent run. A common pattern is triggering an agent after a deployment to automatically generate and publish release notes in your Notion workspace. Connect at Settings → Integrations → Notion.

Agents can now remember things across runs. At the end of a run, Queue extracts key findings — architectural decisions, known bugs, file ownership — and stores them in a vector index. On the next run the agent retrieves relevant memories before starting, giving it background knowledge without burning context window. Enable with memory: persistent in your agent config.

Queue agents can now interact with AWS directly. The integration ships with four built-in tools: aws_s3_read, aws_s3_write, aws_lambda_invoke, and aws_cloudwatch_logs. Authentication uses an IAM role with least-privilege policies — no long-lived access keys are stored. Set up the role and trust relationship from Settings → Integrations → AWS.

The CLI now ships with a built-in eval command. Write a suite of test goals with expected outcomes and run them automatically in CI with queue eval. Compare results across agent config changes before promoting to production. This is the fastest way to catch prompt regressions and tool selection drift early.

Queue agents can now interact with Azure Repos, Boards, and Pipelines from a single run. Use ado_read_file to inspect source code, ado_create_work_item to file bugs and tasks, ado_open_pr to open pull requests, and ado_trigger_pipeline to kick off CI builds immediately after writing code. Authorize via OAuth from Settings → Integrations → Azure DevOps.