May 22, 2026

May 22, 2026

coding_agent

One Sandboxed Runtime to Run Every Coding Agent Your Team Uses

Runtime gives teams a single shared environment for running coding agents like Claude Code, Cursor, Codex, Copilot, and others, with built-in governance, cost tracking, and real-time collaboration. It is part of Y Combinator.

Teams running multiple coding agents hit the same wall: each agent lives in its own silo, nobody can see what it costs, and handing off work mid-session is painful. Runtime is built to fix that.

The product is a shared sandbox layer that sits underneath whichever agents your team already uses. Claude Code, Cursor, Codex, Copilot, Gemini CLI, Devin, and OpenCode are all listed as supported. You configure the environment once: install CLIs, connect APIs and MCP servers, add secrets and custom instructions, set guardrails. Snapshot it, and every new session boots from that snapshot in seconds.

The integrations side covers the tools engineering teams actually touch: Datadog, Salesforce, Stripe, NetSuite, plus monorepos and microservices. The setup is flexible enough that the source describes it as "anything else" beyond the named services. You install from mise, npm, brew, or GitHub.

Beyond environment setup, Runtime lets you build specialized agents tied to specific channels or workflows. The examples in the source are concrete: an alert inspector for an incidents channel, a sales prospector for a revenue channel, a support triager for an inbox. Each agent can run in the background, tag into Slack, Linear, GitHub, or Jira, and post findings without a human prompting it each time.

Collaboration is a first-class feature. Anyone on the team, not just engineers, can prompt an agent and watch it work in real time. Multiple people can be in the same session simultaneously. The output can be a PR, a deploy, a message, a ticket, or a report depending on what the task needs.

Governance is where Runtime makes a case for larger teams. The activity dashboard in the source shows session counts, total cost, cost per task, and per-user attribution. Spend limits, allowlists, and approval gates are described as baked in. Every session exposes tool calls, chain of thought, and file changes in a live view. The dashboard example shows 1,247 sessions and $2,184 in total cost over 30 days, giving you a real sense of the data model.

Access is via browser, terminal, or API. There is a Mission Control CLI and a dashboard at app.runtm.com. Runtime is a Y Combinator company and is currently offering a free tier to start building.

If your team is already using one or more coding agents and you have no shared view of what they are doing or what they cost, this is worth a direct look. The practical first step: install Runtime, snapshot your current dev environment, and run one agent session through it. See whether the visibility layer alone changes how you manage agent work before committing to the deeper integrations.