Key takeaways
- YC-backed (Spring 2026 batch) background coding agent — the company claims teams ship over 30% of pull requests through Replicas
- Pricing is now public: free Hobby tier, Developer at $120/seat/month, Team at $300/seat/month, Enterprise custom — a credits/minutes usage model
- Multi-surface triggers: Linear issues, Slack mentions, GitHub PR comments, web dashboard, or API; sandboxed VMs run full dev environments
FAQ
What is Replicas?
Replicas is a YC-backed background coding agent platform that lets engineering teams delegate tasks from Linear, Slack, or GitHub to AI agents running in sandboxed VMs, producing pull requests for review.
How much does Replicas cost?
As of June 2026, pricing is public: a free Hobby tier (1,200 one-time minutes), Developer at $120/seat/month (5,000 automation minutes), Team at $300/seat/month (15,000 minutes, larger sandboxes), and custom Enterprise plans with SOC 2.
Who competes with Replicas?
Devin (Cognition), Factory, All Hands (OpenHands), Niteshift, Ellipsis, and Codegen. For open-source alternatives, Background Agents (Open-Inspect).
What agents does Replicas use?
Claude Code and Codex, running in isolated sandboxed VMs, with OpenCode support announced as coming soon. Teams can bring their own Claude/Codex credentials.
How do you trigger Replicas?
Mention @tryreplicas in a GitHub PR comment, ping @Replicas in Slack, assign a Linear issue to Replicas, or use the web dashboard or API directly.
Executive Summary
Replicas is a YC-backed background coding agent platform that lets engineering teams delegate coding tasks from their existing tools — Linear, Slack, GitHub, or a web dashboard — to AI agents running in isolated sandboxed VMs.[1] The agents produce pull requests for human review and merge. The company claims teams ship over 30% of their pull requests through the platform, and says it is used by 20+ YC startups.[2] Since our original profile, the company has moved to tryreplicas.com, published seat-based pricing, launched on Product Hunt (June 2026), and named a dozen customers including Mintlify, Composio, and Helicone.[3]
| Attribute | Value |
|---|---|
| Company | Replicas (Replicas Group Inc.) |
| Founded | 2026 (YC Spring 2026 batch; founder Connor Loi) |
| Funding | Y Combinator backed; no additional funding publicly disclosed |
| Headquarters | San Francisco |
| Customers | Mintlify, Composio, Helicone, Dart, Moda, Chronicle Labs, and more |
Product Overview
Replicas positions itself as "the background coding agent" — the key word being background.[1] Unlike foreground tools (Cursor, Claude Code CLI) that require developer attention, Replicas works asynchronously. You delegate a task, go do other work, and come back to a pull request.
The multi-surface trigger model is the key differentiator: any team member can kick off work from wherever they already are — a GitHub PR comment, a Slack message, a Linear issue assignment, or the web dashboard.
The April 2026 YC launch emphasized full local execution inside the sandbox: "A powerful background agent should run everything locally, verify its work, and iterate until it's satisfied."[2] Environments are configurable with files, environment variables, start hooks, warm hooks, and skills/MCPs.
Key Capabilities
| Capability | Description |
|---|---|
| GitHub Integration | Mention @tryreplicas in any PR comment to trigger a fix; also reacts to PR reviews and CI/CD failures |
| Slack Integration | Ping @Replicas in any channel to kick off tasks |
| Linear Integration | Assign issues directly to Replicas to start implementation |
| Web Dashboard | Chat directly with agents for full control |
| Sandboxed VMs | Each agent runs in an isolated VM with its own dev environment |
| Full Dev Environment | Install dependencies, run databases (Redis, Postgres), verify changes locally |
| Configurable Environments | Files, env vars, start hooks, warm hooks, skills/MCPs |
| Custom Workflows | API (POST /v1/replica) for integrating into any workflow |
| Multiple Agents | Claude Code and Codex in parallel; OpenCode support announced as coming soon |
Product Surfaces
| Surface | Description | Availability |
|---|---|---|
| Web Dashboard | Direct chat and task management | GA |
| GitHub Bot | @tryreplicas mention in PR comments | GA |
| Slack Bot | @Replicas mention in channels | GA |
| Linear Integration | Issue assignment trigger | GA |
| API | REST API for custom workflows | GA |
Technical Architecture
Each task creates an isolated sandboxed VM with a full development environment.[1] The VM includes the project repository, installed dependencies, running databases and services, and a coding agent (Claude Code or Codex). Changes are verified locally within the sandbox before being submitted as a pull request. Teams can bring their own Claude/Codex credentials rather than being locked to a single model vendor.[2]
Key Technical Details
| Aspect | Detail |
|---|---|
| Deployment | Cloud-hosted (SaaS) |
| Runtime | Sandboxed VMs per task (Team plan: 4 vCPU, 16 GB RAM, 32 GB disk) |
| Model(s) | Claude Code, Codex (OpenCode announced) |
| Integrations | GitHub, Slack, Linear, REST API |
| Open Source | No (commercial SaaS) |
Strengths
- Multi-surface triggers — Meet developers where they are: GitHub PR comments, Slack, Linear, dashboard, or API; lowest friction in the category[1]
- Transparent public pricing — Free tier plus $120 and $300 seat tiers removed the earlier "contact sales" opacity[3]
- YC backing — Signals credibility, access to network, and funding runway[4]
- Sandboxed full dev environments — Not just code generation; agents can run databases, install dependencies, and verify changes end-to-end
- Team-oriented — Designed for engineering teams, not solo developers; anyone can assign work
- Custom workflow API — REST API enables programmatic integration into CI/CD, internal tools, and custom automation
- Adoption signals — "Used by 20+ YC startups," with named customers including Mintlify, Composio, and Helicone; the 30%+ PR throughput claim is the company's own[2]
Cautions
- Solo-founder risk — The YC profile lists a team size of one (founder Connor Loi); key-person and support-capacity risk for teams making it a core workflow dependency[4]
- Unverified adoption claim — The "30%+ of pull requests" figure is the company's own marketing copy; no independent case studies or benchmarks back it[1]
- Closed source — No transparency into how agents work or handle sensitive code
- Usage-minute billing complexity — Credits/minutes on top of seats makes cost forecasting harder; automation minutes are capped on Developer and Team tiers[3]
- VM overhead — Spinning up full dev environments per task adds latency vs lightweight sandbox approaches
- Competitive market — Devin, Factory, Codegen, and open-source alternatives all competing for the same budget
What Developers Say
Reactions from the June 3, 2026 Product Hunt launch (no Hacker News discussion of Replicas was found as of June 2026):[5]
"We use replicas in our day-to-day work to develop multiple features in parallel... Cloud agents are the next big thing in the AI coding world, and Replicas is the best one out there right now." — Ivan Zvonar (Dart), Product Hunt
"Spawn Claude Code in a VM with tooling ready to go, this is genuinely what I keep wishing existed when delegating background tasks." — Ergin Murat, Product Hunt
"The cloud VM approach makes sense especially when multiple agents need to run in the background without depending on someone's laptop." — Ada Johnsen, Product Hunt
Note that launch-day Product Hunt comments skew promotional; there is little independent long-term usage feedback yet.
Pricing & Licensing
Pricing became public after our original profile.[3]
| Tier | Price | Includes |
|---|---|---|
| Hobby | Free | 1,200 minutes of human-initiated workspace usage + 1,200 API/automation minutes (one-time); 2 automations, 3 repos, 5 environments |
| Developer | $120/seat/month | Unlimited human-initiated workspaces; 5,000 automation/API minutes/month; 10 repos, 15 environments, 5 automations; API access |
| Team | $300/seat/month | Unlimited human-initiated workspaces; 15,000 automation/API minutes/month; unlimited repos and automations; larger sandboxes (4 vCPU, 16 GB RAM); shared Slack support |
| Enterprise | Custom | Unlimited usage and automations, custom API rates, SOC 2 |
Licensing model: Commercial SaaS. Credits are acquired via subscription or direct purchase, then spent on usage; API workspaces bill separately at period end.
Hidden costs: Automation/API minutes beyond included allotments; VM compute is baked into the minutes model.
Competitive Positioning
Direct Competitors
| Competitor | Differentiation |
|---|---|
| Devin | Full autonomous engineer with browser/terminal — Replicas focuses on task delegation with human review |
| Factory | Enterprise-focused with compliance features — Replicas is lighter, more developer-friendly |
| All Hands (OpenHands) | Open source — Replicas is commercial with tighter integrations |
| Codegen | Code generation API — Replicas is full background agent with dev environments |
| Background Agents (Open-Inspect) | Open-source, self-hosted — Replicas is managed SaaS |
| Tembo | Orchestration platform (BYOK, local agents) — Replicas is cloud-hosted background agents |
When to Choose Replicas Over Alternatives
- Choose Replicas when: Your team wants to delegate tasks from Linear/Slack/GitHub without managing infrastructure
- Choose Devin when: You need a more autonomous agent that can browse the web and handle complex multi-step tasks
- Choose Background Agents when: You want open-source, self-hosted background agents
- Choose Tembo when: You want to orchestrate your own agents with your own keys
Ideal Customer Profile
Best fit:
- Engineering teams (5-50 engineers) with Linear/GitHub/Slack workflows
- Teams wanting to offload routine coding tasks (bug fixes, small features, test writing)
- Organizations comfortable with cloud-hosted code execution
Poor fit:
- Solo developers (overkill; use Claude Code or Cursor directly — though the free Hobby tier lowers the barrier to trying it)
- Teams with strict data residency or air-gapped requirements
- Organizations needing fully autonomous agents without human review
Viability Assessment
| Factor | Assessment |
|---|---|
| Financial Health | Early — YC-backed (Spring 2026), no additional funding publicly disclosed |
| Market Position | Early but credible — 20+ YC startup customers, named logos (Mintlify, Composio, Helicone) |
| Innovation Pace | Active — YC launch (April 2026), Product Hunt launch (June 2026), public pricing, OpenCode support announced |
| Community/Ecosystem | Limited — Product Hunt buzz but no Hacker News footprint or public community as of June 2026 |
| Long-term Outlook | Cautiously positive — fast shipping pace, but a solo-founder company in a crowded category |
The background coding agent market is heating up rapidly (Devin, Factory, Codegen, etc.). Replicas' multi-surface trigger model (Linear + Slack + GitHub + API) is a genuine differentiator — most competitors only support one or two trigger surfaces.
Bottom Line
Replicas occupies an interesting position in the background coding agents category: it's not trying to be a fully autonomous engineer (like Devin) or an open-source framework (like OpenHands). Instead, it's a managed service that connects your existing workflow tools (Linear, Slack, GitHub) to sandboxed coding agents. The multi-surface trigger model is the strongest version of this pattern we've seen, and the move to transparent pricing and named customers since March 2026 strengthens the case.
Recommended for: Engineering teams wanting to delegate routine coding tasks from their existing Linear/Slack/GitHub workflows.
Not recommended for: Teams needing data residency, organizations wanting open-source/self-hosted solutions, or teams uncomfortable betting a core workflow on a solo-founder startup.
Outlook: The background coding agent category is crowded and consolidating fast. Replicas is shipping quickly — public pricing, Product Hunt launch, OpenCode support announced — and YC backing gives it runway, but it will need to grow beyond a one-person team and validate its throughput claims as competitors mature.
Research by Ry Walker Research • methodology