Replicas
A background coding agent platform that delegates software tasks to AI agents running in sandboxed development environments, producing pull requests from GitHub, Slack, Linear, or API triggers.
At a Glance
About Replicas
Replicas is a Y Combinator-backed (Batch P26) coding agent platform that lets engineering teams delegate software tasks to AI agents working inside isolated virtual machines. The platform is currently announcing its V1 launch and is accessible via web dashboard, CLI, and API. Agents run Claude Code or Codex, produce pull requests, and can be triggered from GitHub, Slack, Linear, or directly through the Replicas API.
What It Is
Replicas sits in the background coding agent category — a class of tools where AI agents autonomously implement code changes rather than just suggesting them inline. A developer assigns a task (a bug fix, a feature, a refactor), and an agent picks it up, works inside a sandboxed VM with its own dependencies and databases, and opens a pull request for human review. The human reviews and merges; the agent handles the implementation work.
How the Workflow Operates
The three-step flow the product describes is: delegate, sandbox, pull request.
- Delegate: Tasks come in from Linear issue assignment, a Slack @mention, a GitHub PR comment mentioning @tryreplicas, or a direct chat in the web dashboard.
- Sandbox: Each agent runs in an isolated VM. Dependencies can be installed, databases (Redis, Postgres) can be spun up, and the agent verifies changes locally before opening a PR.
- Pull Request: The resulting PR is ready for human review and merge. No code is stored in Replicas' databases; only the connected GitHub repository holds the code.
Supported Agents and Integrations
Replicas currently supports Claude Code and Codex as the underlying coding agents, with OpenCode support listed as coming soon. Integration touchpoints include:
- GitHub: Mention @tryreplicas in any PR comment to trigger a fix
- Slack: Ping @Replicas in any channel to start a task
- Linear: Assign issues directly to Replicas
- Dashboard: Full chat interface at tryreplicas.com/dashboard
- API: POST to
/v1/replicawith repository, message, and agent selection for custom workflow automation
Security and Data Model
The platform states that each workspace runs in an isolated VM, code never leaves the connected repository, and Replicas does not train on user data. Only the user has SSH access to their workspaces. Sleeping sandboxes are archived but not read, and they are automatically deleted after 7 days of inactivity. No code is stored in Replicas' databases.
Billing and Usage Model
Replicas uses a credit/minutes model layered on top of per-seat subscriptions. Human-initiated workspace usage (dashboard, Slack, Linear, GitHub) is covered by seats. API and automation workspaces are billed separately by compute minutes at period end. Workspaces automatically sleep when idle and wake on demand, so users pay only for active compute time.
Current Status
The homepage announces Replicas V1 and references a Product Hunt launch. The company is backed by Y Combinator (P26) and lists a careers page on YC. The vendor's homepage claims engineering teams use Replicas to ship over 30% of pull requests, attributing this to their listed customer logos.
Community Discussions
Be the first to start a conversation about Replicas
Share your experience with Replicas, ask questions, or help others learn from your insights.
Pricing
Hobby
Free tier with one-time minutes allocation for getting started.
- 1,200 minutes of human-initiated workspace usage (one-time)
- 1,200 minutes of API + automation usage (one-time)
- Automations (limited to 2)
- Warm pools and warm hooks
- Up to 3 repositories
Developer
Per-seat plan for individual developers with unlimited human-initiated workspaces.
- Unlimited human-initiated workspaces
- 5,000 included automation/API minutes per month
- Up to 10 repositories
- Up to 15 environments
- Up to 5 automations
- Warm pools and warm hooks
- API access
Team
Per-seat plan for teams with higher limits, shared Slack support, and upgraded sandbox resources.
- Unlimited human-initiated workspaces
- 15,000 included automation/API minutes per month
- Unlimited repositories
- Unlimited automations
- Higher API rate limits
- Warm pools and warm hooks
- Auto-upgraded sandbox resources (4 vCPU, 32 GB disk, 16 GB memory)
- Shared Slack support channel
Enterprise
Custom plan with unlimited usage, SOC 2, and custom rate limits.
- Unlimited usage
- Unlimited automations
- Custom API rates
- Custom rate limits
- Custom warm hooks and pools
- Shared Slack support channel
- SOC 2
Capabilities
Key Features
- Background coding agents that work autonomously in sandboxed VMs
- Claude Code and Codex support (OpenCode coming soon)
- GitHub PR comment trigger via @tryreplicas mention
- Slack task delegation via @Replicas mention
- Linear issue assignment to Replicas agent
- Web dashboard with chat, diff, and plan views
- REST API for custom workflow automation
- Isolated VM per agent with full dev environment (dependencies, databases)
- Automatic workspace sleep and wake on demand
- No code stored in Replicas databases
- SSH access restricted to user only
- Warm pools and warm hooks
- Auto-upgraded sandbox resources on higher tiers
- Unlimited human-initiated workspaces on paid plans
- Shared Slack support channel on Team and Enterprise