Runtime

Runtime is a shared runtime layer for coding agents, designed to let teams across an organization run sandboxed, governed agents against real systems without building their own infrastructure. Instead of every team plumbing together its own isolated bot, Runtime plugs into popular coding agents—Claude Code, Cursor, Copilot, Gemini CLI, Devin, OpenCode, and others—and supplies the environments, policies, orchestration, and observability needed to ship pull requests, tickets, reports, and more safely and consistently.

• Multi‑agent coding runtime: Works with a wide range of popular coding agents while Runtime handles environments, policies, and orchestration, so teams can swap models or tools without re‑plumbing anything.
• Sandboxed compute with snapshots: Pre‑warmed VMs mirror monorepos and microservices, with custom CLIs, APIs, and secrets pre‑configured. Sessions start quickly and remain isolated from production.
• Background and interactive agents: Supports both live interactive sessions and autonomous background agents that pick up tickets, issues, or Slack messages, then write code, run tests, and return pull requests.
• Mission Control & observability: A central dashboard to monitor every agent session—tool calls, chain of thought, file diffs, costs, and approval status—plus spend limits and allowlists.
• Deep SaaS and data integrations: Comes with connectors for data warehouses, billing, HR, CRM, support, alerting, and engineering tools, and can call anything reachable via API, CLI, SDK, or MCP.
• Self‑hostable and open source pieces: Can run fully inside a company’s cloud with its own models and storage, backed by open source components for the CLI, API worker, and environment templates.
• Model and tool flexibility: Companies can bring their own model keys and tools instead of being locked into a single agent vendor, allowing use of preferred LLMs and custom integrations.

1. Sign up and set up your organization – Create a Runtime account, then configure your team structure, role‑based access, and initial security settings.
2. Connect your coding agents – Bring your own model keys (e.g., Claude, OpenAI) and register the agents you already use (Claude Code, Cursor, Copilot, etc.) within Runtime.
3. Create sandbox environments – Define templates that mirror your repos, microservices, and secrets, so every agent session starts in a pre‑warmed, consistent VM.
4. Define guardrails and policies – Set approval gates, spend limits, file‑access rules, and allowlists to keep agents away from sensitive data and risky side effects.
5. Launch agents from your tools – Trigger agents via Slack, Linear, GitHub, Jira, or command line; they can run interactively or autonomously, pick up work, and return pull requests or tickets.
6. Monitor and iterate in Mission Control – Use the central dashboard to track tool calls, diffs, costs, and session logs, then refine policies as you scale.

Runtime offers a Freemium model with plans that scale from individual exploration to full enterprise deployment. All plans include sandboxed sessions, core observability, and access to popular coding agents; higher tiers unlock parallel sessions, collaboration, advanced governance, and self‑hosting options.

• Free – $0/month: 500 credits, 1 session, Claude Code, Codex, OpenCode, basic sandbox, live preview/terminal, GitHub import/export, community support.
• Builder – $29/month: 10 parallel sessions, all coding agents, standard sandbox, live preview/terminal, GitHub import/export, custom templates, BYOK, SDK/CLI/API access, priority support.
• Teams – Starting at $99/seat/month: Unlimited sessions, all coding agents, real‑time collaboration, shared templates/secrets, activity dashboard, role‑based access, spend controls, BYOK, SDK/CLI/API access, priority support.
• Enterprise – Custom pricing: Everything in Teams, plus self‑hosted runtime, bring‑your‑own sandbox provider, environment templates, custom sandbox resources, dedicated infrastructure, SSO/SAML/SCIM, compliance features, and a dedicated support engineer.

Disclaimer: Pricing information may not be up to date. For the most accurate and current details, refer to the official Runtime website.

• Web dashboard (Mission Control): Full observability session cockpit accessible via any modern browser.
• CLI and API: Command‑line interface and REST API for managing sandboxes, launching agents, and integrating with existing pipelines.
• SDK: Programmatic access for building custom tooling and embedding the runtime into internal platforms.
• Self‑hosted option: Deploy the entire runtime inside your own cloud with your own compute and storage (Enterprise plan).
• Integrations: Native connectors for Slack, Linear, GitHub, Jira, data warehouses, billing, CRM, and more; reachable via MCP, APIs, or SDKs.

• Start with a narrow scope: Pilot Runtime with a single team and a limited set of agents before rolling it out across the organization.
• Pre‑configure robust sandbox templates: Invest time early in baking in dependencies, CLIs, and secrets so sessions start instantly and consistently.
• Set clear cost and access boundaries early: Use spend limits and file rules to avoid unexpected bills and accidental exposure of sensitive data.
• Encourage session sharing for collaboration: Let multiple teammates step into the same agent session to review code, hand off work, or debug together in real time.
• Review observability data regularly: Check Mission Control not just for anomalies, but to understand where agents save time and where prompts need tuning.
• Treat agent policies as living documents: As teams discover new use cases, update guardrails, approved tools, and model choices to match evolving needs.

• Platform and DevOps engineers who need to offer a secure, shared agent infrastructure across the company without every team building its own sandboxes and orchestration.
• Software engineering and product teams that want coding agents to tackle real coding tasks—bug fixes, feature prototypes, PRs—while maintaining control over code review and environment integrity.
• Customer support leaders looking to automate ticket triage and response drafting with guardrails that prevent data leakage and ensure quality.
• Finance and operations professionals who need to automate data pulls, reconciliations, and internal reports from multiple systems in a governed way.
• Design and marketing teams that want to experiment with copy, page layouts, or asset generation using agents that touch live content, but only inside safe sandboxes.
• Compliance, audit, and risk managers exploring automated control checks or policy enforcement without giving agents unfettered access to production.