Hoplite
Also known as: Hoplite, hoplite.sh, Carbon Copy Markets
Cloud coding agents in isolated sandboxes that import your local Claude Code, Codex or OpenCode setup, run in parallel with subagents, start from Slack, Linear, Sentry or a schedule, and hand back a pull request to review.
Hoplite runs coding agents in isolated cloud sandboxes and makes the pull request the interface. A team connects a GitHub repository and imports its local setup from Claude Code, Codex or OpenCode, including session history, skills, MCP servers and the command line tools a project needs. Each thread then gets its own machine, on Modal by default, with dependencies installed, tests running and the app booted on a live preview URL. The cloud engine is Hoplite's own agent over the model a customer picks, a cloud Claude Code engine is switched on per organization, and the Mac desktop app supervises local Claude Code and Codex runs.
Threads run in parallel, so a whole backlog can be fanned out at once and reviewed as each one finishes, and a thread can delegate to subagents, 8 on paid plans and 4 on Free. Work starts by hand in the web app or the desktop app, from a Slack mention, from iMessage, from a Linear delegation or from an @usehoplite comment on a pull request.
Automations start work without anyone asking. They fire on a schedule or a cron expression, on GitHub pull request events and failed CI checks, on new or regressed Sentry issues, on Linear and Slack events, when another thread finishes, or on a request to a webhook URL. Templates ship ready to switch on, covering a daily digest, issue triage, dependency patrol and Sentry alert triage. A separate PR Autofix loop, on by default, watches checks and review threads and keeps pushing fixes to the same branch until everything is green, while auto merge stays off unless a team turns it on.
Every thread is shared with the workspace, so teammates can watch a run and steer it. Since mid October cloud runs approve their own tool calls within the command policy, protected paths and workspace policy an admin sets, while local runs on a developer's machine keep the approval flow. Agents cannot mark their own preview checklist as reviewed, and the merge decision stays with the pull request review. An activity timeline records each operation the agent called and the result it received, and the agent keeps personal memories per user that the user can review, edit or delete.
Workspace roles run from owner and admin through member and viewer, with two factor sign in, passkeys and domain based auto join, and changing an MCP server needs a fresh owner or admin check. MCP credentials are encrypted individually with AWS KMS, and repository access is brokered through short lived, repository scoped GitHub installation tokens. There is no single sign on, SOC 2 is in progress, and no report is published.
Customers pick the model per thread, project or automation from Anthropic, OpenAI, Google, Mistral, DeepSeek, Kimi, GLM and OpenRouter models, bring their own keys, or connect their ChatGPT, Claude or Grok subscriptions with failover between them. A public API with an OpenAPI specification, an OAuth protected MCP server and an npm command line client let other systems and agents start and inspect threads. Hoplite is built by Carbon Copy Markets, Inc., a Y Combinator company.
From our cloud coding agents video
Devin, Factory, Jules, OpenHands: What Each Cloud Coding Agent Does Best
Our video on the 11 cloud coding agents in the index names what each one does best. For Hoplite, that is the handoff from your laptop to the cloud.
The player starts at Hoplite's part, 6:46 in. Recorded in October 2026; the profile here is live and updates as research is verified.
Vendor details
Canonical URL
https://hoplite.sh
Category
Coding agent
Subcategory
Cloud coding agent platform with parallel sandboxes and a pull request workflow
Funding status
Y Combinator, Summer 2026 batch; no priced round disclosed
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
A GitHub App connects repositories, pull requests, checks and reviews (GitLab is not supported). Linear brings in issues and delegations, Slack and iMessage start and steer work, and Sentry feeds error triage and takes status updates. Hoplite also connects to Sazabi and Axiom, and up to sixteen remote MCP servers over HTTP or SSE can be added per project. Inbound webhooks start automations, and a public API with service accounts and signed webhooks, an MCP server at api.hoplite.sh/mcp and the @usehoplite/cli npm client let outside systems drive the platform.
In practice
A new Sentry regression lands overnight. An automation opens a thread for it, the agent runs a root cause pass, writes a failing test, fixes the bug and opens a pull request, and updates the Sentry status for the morning review.
A founder has twelve small backlog tickets. Each goes to its own Hoplite thread in parallel, and every agent checks its change against a live preview of the app before handing back a pull request.
An engineer is midway through a Claude Code session when it is time to leave. The command line client hands the conversation to a cloud thread that keeps working, and a teammate picks it up in the shared workspace.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
11.5 / 14 capabilities · 82%
| Integrations & Tool Calling | Full |
|---|---|
|
Native GitHub tools open, update, review and merge pull requests through repository scoped installation tokens (GitLab is not supported), Sentry comments and status changes run on the integration's write capability, Linear and Slack start and steer work, iMessage, Sazabi and Axiom are connected, and up to sixteen remote MCP servers over HTTP or SSE can be added per project with their own credentials, with no stdio servers. Sourcehoplite.sh/docs/integrations, docs/agent/mcp and docs/agent/toolsread 2026-10-28 |
|
| Workflow Orchestration | Full |
|
Threads run in parallel, one sandbox each (5 at once on Free, 50 per seat on Team), a parent run delegates to subagents (8 on paid plans, 4 on Free) that share its workspace or get their own, and since 19 October delegated agents can use other providers' models. Swarms run for enabled organizations, and automations chain when one thread finishing starts another, capped at 16 hops with repeat events suppressed. The cloud engine is Hoplite's own agent over many models; a cloud Claude Code engine is enabled per organization, and the desktop app supervises local Claude Code and Codex runs. Sourcehoplite.sh/changelog 7, 13, 14 and 19 October 2026, docs/agent/tools and docs/desktopread 2026-10-28 |
|
| Knowledge Grounding & RAG | Partial |
|
Context is assembled for each run: each thread works from a checkout of the connected GitHub repository, AGENTS.md (with CLAUDE.md as a fallback), instructions and skills are read from disk, and agents can search thread history within the project. There is no persistent index, graph or embeddings layer over the codebase or the customer's knowledge. Sourcehoplite.sh/docs/agentread 2026-10-28 |
|
| Human Oversight & Guardrails | Partial |
|
Since mid October cloud agents automatically approve tool calls, including connected service writes, so a cloud run does not pause for a tool approval and the docs describe no way to require one; repository scope, command policy, protected paths and workspace policy still constrain what a run may do. Local CLI and desktop runs on the user's own machine keep the approval flow. PR Autofix is on by default and auto merge is a separate setting that is off, so the merge decision stays with the customer's pull request review, and agents cannot mark their own preview checklist as reviewed. The homepage FAQ, the multiplayer page and a September blog post still describe approval gating. Sourcehoplite.sh/docs/agent/tools, docs/github and changelog 11 and 13 October 2026read 2026-10-28 |
|
| Security, Identity & Governance | Partial |
|
Access is managed through workspace roles of owner, admin, member and viewer, two factor sign in and passkeys, automatic workspace join by verified DNS domain, a fresh owner or admin check on any MCP server change, and scoped API keys and service accounts; MCP credentials are encrypted individually with AWS KMS. There is no SSO, SAML or SCIM (the docs have no SSO page). Hoplite says SOC 2 is in progress, and no report or trust page is published. Sourcehoplite.sh/docs/workspace, docs/workspace/security and hoplite.shread 2026-10-28 |
|
| Observability & Auditability | Full |
|
The activity timeline shows each operation the agent called and the result it received, every thread streams live and stays visible to the workspace, automations keep a run history with each run's thread, status and duration, transcripts can be exported, usage is exported as CSV, and a checkpoint restore API rolls a thread back. A settings audit log is mentioned on the GitHub docs page. Retention is not stated. Sourcehoplite.sh/docs/agent/tools, docs/automations, docs/github and docs/apiread 2026-10-28 |
|
| Memory & State Persistence | Full |
|
Personal memories are a documented store the agent writes: it can save or update a concise, stable cross thread preference without per write approval, and Hoplite may curate a preference stated in an opening message. The scope is per user, available only to agents that user starts, and the user reviews, creates, edits and deletes memories under Settings, Agent personalization, which is the purge path. Unattended automation runs with no initiating user do not inherit anyone's memories. A cloud thread also keeps its own sandbox state across runs, and local Claude Code, Codex or OpenCode session history can be imported. Sourcehoplite.sh/docs/agent (personal memories)read 2026-10-28 |
|
| Deployment & Data Residency | Partial |
|
Hoplite runs in managed cloud sandboxes on Modal by default, with Daytona and a Docker Compose VM as alternatives, with no region choice, VPC or self hosted option (the docs have no self host page). The Enterprise tier offers private networking and retention in one line with no named mechanism. The Mac desktop app runs local threads against a folder on the user's machine through local providers, using the customer's own agent. Sourcehoplite.sh/docs/sandboxes, hoplite.sh/pricing and docs/desktopread 2026-10-28 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Automation templates switch on in one click, covering a daily digest, issue triage, dependency patrol and Sentry alert triage, which runs a root cause pass, fixes the issue, opens a pull request and updates the Sentry status. Skills are defined at project, organization and personal scope, and an MCP directory lists connectable servers. There is no marketplace. Sourcehoplite.sh/product/automations, docs/agent and docs/agent/mcpread 2026-10-28 |
|
| Triggers & Channel Coverage | Full |
|
Automations start on schedules, by preset, interval or cron in a chosen timezone with a floor of 60 minutes on Free and 5 on paid plans, and on events: GitHub pull request changes and failed CI checks, new or regressed Sentry issues, Linear issue changes, Slack messages or reactions, a finished Hoplite thread, and a per automation webhook URL, with filters on event data. Threads also start from a Linear delegation, an @usehoplite mention on a pull request, a Slack mention, iMessage or the API. Sourcehoplite.sh/docs/automations and product/automationsread 2026-10-28 |
|
| Model Flexibility & Routing | Full |
|
Customers choose the model per thread or per message, and a default per project and per automation, from Anthropic, OpenAI, Google, Mistral, DeepSeek, Kimi, GLM and OpenRouter models. They can bring their own Anthropic or OpenAI keys or a custom gateway, and connect their own ChatGPT, Claude or Grok subscriptions, with failover between them and several accounts per provider since late October. Sourcehoplite.sh/docs/agent/models and changelog 27 and 28 October 2026read 2026-10-28 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Hoplite has a public REST API with an OpenAPI specification, scoped API keys and service accounts, and signed webhooks for building a budgeted task pipeline, along with an OAuth protected streamable HTTP MCP server at api.hoplite.sh/mcp (connected with hoplite mcp connect since 26 October) and the @usehoplite/cli npm client. A Platform API is a release candidate for a limited set of organizations, and no SDK is published. Sourcehoplite.sh/docs/api, docs/agent/mcp and changelog 26 October 2026read 2026-10-28 |
|
| Testing, Debugging & Optimization | Partial |
|
The agent runs the project's own test suite, checks its change against a managed preview checklist and can record video of new features, a /goal evaluator judges whether a thread met its goal, and the PR Autofix loop iterates until CI checks pass, all using checks the customer already owns. There is no evaluation harness, scored test cases or optimization loop for the agent itself, and a Hoplite blog post on evaluating an agent over 30 backlog tickets describes a study a customer could run, not a product feature. Sourcehoplite.sh/product/pr-autofix, docs/agent and docs/agent/toolsread 2026-10-28 |
|
| Browser & Computer Use | Full |
|
Browser automation runs through the sandbox shell with bundled Chrome and Playwright, and the agent drives it against a managed preview of the running app to verify its own change, take screenshots and complete preview checklists, inside a sandbox Hoplite operates. There is no desktop computer use. Sourcehoplite.sh/docs/agent/tools, docs/sandboxes and blog/verify-agent-work-with-tests-and-a-browserread 2026-10-28 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Team workspaces now allow 150 concurrent sandboxes per seat, 500 new threads per hour, 100 active automations on schedules as often as every minute and 500 iMessages per day. Existing Team subscriptions get the new limits automatically. Workspaces that move to Plus from now on run sandboxes of up to 8 vCPU and 32 GiB RAM, while those already on Plus keep 16 vCPU and 64 GiB.
Bears on: Pricing / packaging
View sourceCommits from a Hoplite thread can now be authored as the GitHub account of the person who started it, with pull requests opened as that person and Hoplite credited as coauthor. It is on by default for new workspaces, and owners of existing workspaces can switch it on. Automations still commit as Hoplite.
Bears on: Observability / auditability
View sourceA new /loop command repeats a prompt on a set interval inside a cloud thread or swarm, so an agent can recheck a deployment or review progress every few minutes in the same conversation. Loops keep running while the browser is closed, wait while the conversation is busy and expire after seven days.
Bears on: Workflow orchestration
View sourcePricing
Team at $60 per seat a month, or $50 billed yearly. Free plan at $0. A single seat Plus plan at $20 a month is rolling out, per the changelog.
Seats, each with a monthly usage allowance consumed by model and sandbox use. Each automation can carry an optional spend limit.
Included quota
Free covers one user, 5 concurrent 2 CPU and 8 GB cloud sessions, 5 projects, any platform model and an undisclosed monthly credit allowance with a daily cap. Plus is a single seat with up to 15 projects and 300 credits a month, charged up front with no trial. Team adds 50 concurrent sandboxes per seat, unlimited projects, 250 new threads an hour, bigger sandboxes, usage analytics and more included usage per seat (500 credits per the changelog).
What is public
Hoplite publishes three plans on its pricing page. Free is $0 with a card required, for one user with 5 concurrent sessions, 5 projects and any platform model. Team is $60 per seat a month, or $50 billed yearly, with 50 concurrent sandboxes per seat, unlimited projects, 250 new threads an hour and more included usage. Enterprise is custom and annual. The 4 October changelog adds Plus, a single seat plan at $20 a month or $200 a year with 300 credits, rolling out gradually and still absent from the pricing page cards, though the home FAQ says upgrade to Plus or Team and the billing docs sell it with no trial. The 12 October changelog records that Team stopped offering a free trial, and the 7 October entry prices sandboxes at $0.08 to $1.50 an hour with a credit worth $1.
Billing mechanics
Per seat subscription, monthly or yearly, with included usage per seat. Automations can carry their own spend limit, and the Slack and iMessage assistant's own replies are paid by Hoplite under a daily fair use limit while the coding threads it starts use workspace credits.
Cost watchouts
Included usage is the main unit of cost, and what a credit buys is not published. Hoplite offers bring your own keys, which would move inference cost to the customer's provider bill. Custom security, private networking, contractual SLAs and architecture review sit in the annual Enterprise tier.
Variable cost rationale
Seats are a fixed monthly price. Agent work, meaning model calls and sandbox use, draws down each seat's included usage, and what a credit buys is not published. Heavier workloads use it up faster. No overage rate is published. When the workspace runs out of credits, automations back off for an hour, so work stops rather than billing on. Spend limits on each automation and workspace spending caps limit usage.
Additional watchouts
Cost cannot be modeled past the seat price because what a credit buys and any overage are unpublished. The Free plan requires a card.
Overage / add-ons
A credit is $1, bought prepaid, and sandboxes draw $0.08 to $1.50 an hour while they work; bring your own keys and connected subscriptions use no credits for model calls. No overage rate is published. Hoplite has credit holds and spending caps, and automations back off for an hour when the workspace is out of credits, so work stops rather than an overage rate applying.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free plan at $0 for one user, with a card required and nothing charged. Team no longer offers a free trial.
Lowest paid plan
Team at $60 per seat per month, or $50 billed yearly, on the pricing page. Plus at $20 a month is rolling out per the 4 October 2026 changelog.
Commercial notes
Free and Team have self serve signup, and Enterprise is booked by email to sales. Plans can be switched or canceled at any time.
Key ambiguities
What a credit buys in model use is not stated, and the pricing page still names Plus without a card or price, which appears only in the changelog and billing docs.
Cancellation / refund
Plans can be switched or canceled at any time. Enterprise is an annual agreement.
Support SLA / resale
Team includes standard support. Enterprise adds contractual SLAs and an architecture review.
Missing data
What a credit buys, overage rate, the baseline behind 2x usage, Enterprise floor, and whether bring your own keys usage is metered.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- 10Web — Agentic website platform whose AI agents build, host and manage…
- AgentUI — Managed AI app builder for operations teams that generates and hosts…
- Aider — Open source, model agnostic terminal coding agent that edits your…
- Amp — Amp is a coding agent from Amp Frontier Corporation that runs each…
- Anthropic Claude Code — Anthropic's agentic coding system across terminal, desktop, IDE, web…
Alternatives to Hoplite
The closest documented capability profiles to Hoplite among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Hermes Agent12.5 / 14Fuller documented coverage on Human Oversight & Guardrails and Deployment & Data Residency
- Anything11.0 / 14Fuller documented coverage on Testing, Debugging & Optimization
- Base4413.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Security, Identity & Governance
- Blink10.0 / 14A lighter documented profile than Hoplite
- Cline12.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Deployment & Data Residency
- JetBrains AI12.0 / 14Fuller documented coverage on Human Oversight & Guardrails and Deployment & Data Residency
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded