June Product Update

Here's everything Plural shipped in June.

🔧Workbenches Enhancements: Cloud-to-local coding agents, codebase-memory MCP, and usage metrics

Cloud-to-local coding agents. AgentRun sessions can now move between cloud and local execution, with the harness generating patches, session archives, and downloadable artifacts on each handoff.
Codebase-memory MCP server. A new MCP server gives the agent harness durable codebase memory, wired into the analyze, babysit, and write system prompts so agents carry context across steps.
Usage metrics and accounting. Workbenches now surface usage metrics in the UI, with workbench usage accounting and token reporting behind it. Flows workbenches also gained search.
Per-run model selection. Workbenches let you choose and manage the model per run, with improved model routing logic and a prompt-mode dropdown.
Git import and local run. New "import git" flow in workbench setup and a local run button to kick off a run against your own checkout.

Terragrunt joins the supported toolchains, and Terraform runs are now faster and safer thanks to plan and state caching.

Terragrunt support. InfrastructureStack can now drive Terragrunt projects alongside Terraform, Ansible, and the rest.
Cached Terraform plan and state. Plans and state are cached between runs, cutting redundant work and reducing churn.
Targeted GVKs. Stacks can target a specific group/version/kind, narrowing what a run reconciles.
Copy outputs. A copy button on StackOutput makes it one click to grab a stack's outputs.
Kustomize renderer refactor. The kustomize rendering path was refactored for reliability.

We're currently attending AWS Summit in DC! Come by booth #477 and say hi!

We had a great time at AWS Summit in NYC earlier this month!