Warp · 索引內容

{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"What counts as a credit in Warp?","acceptedAnswer":{"@type":"Answer","text":"Credits are Warp's unit for agent requests. You use credits when you run Warp Agent in the app or CLI, or run Oz cloud agents. Build includes 1,500 credits ($20) each month, Business includes 1,500 credits ($20) per seat, and Max includes 18,000 credits ($240). How many credits an action costs depends on the model, the size of the task, how much context and codebase are involved, and whether the agent runs locally or in the cloud. Cloud and managed-agent runs can also include platform and compute costs. Learn more about how credits work in our docs."}},{"@type":"Question","name":"Is the Warp Agent CLI included in my plan?","acceptedAnswer":{"@type":"Answer","text":"Yes. The Warp Agent CLI is included with every Warp plan on macOS and Linux. Sign in with the same Warp account to use your plan, credits, models, and synced conversations. CLI requests use the same credits as the Warp app. Free users can buy credits or add an OpenAI, Anthropic, or Google API key or SuperGrok subscription with /api-keys. Warp's Auto mode…

{"@context":"https://schema.org","@type":"Blog","@id":"https://www.warp.dev/blog#blog","name":"Warp Blog","url":"https://www.warp.dev/blog","description":"Engineering deep-dives, product launches, and how 800,000 developers use Warp to build with AI agents.","publisher":{"@id":"https://www.warp.dev/#organization"},"blogPost":[{"@type":"BlogPosting","headline":"The Factory Stack","url":"https://www.warp.dev/blog/the-factory-stack","datePublished":"2026-09-05T15:00:00.000Z","description":"Software factories should be built on an infrastructure stack that is open, composable and defined in code."},{"@type":"BlogPosting","headline":"Introducing Factory Benchmarks","url":"https://www.warp.dev/blog/warp-factory-benchmarks","datePublished":"2026-09-03T15:00:00.000Z","description":"Today we are launching Benchmarks in Warp Factories early access: agent benchmarks on your own code that drive direct improvements in software factory performance. Rather than guessing at what model and skill configurations are best, you can now easily test, measure, and put your findings into practice."},{"@type":"BlogPosting","headline":"Closing the loop with self-improving cloud software factories","url":"htt…

EngineeringAdopting the software factory model: crawl, walk, runZach Lloyd|September 15, 2026The software factory approach (a closed agentic loop that runs in the cloud) is growing in popularity, but it can be daunting to adopt. In this post I’ll go through “crawl, walk, run” steps for making the transition from local, interactive agents to automated cloud development.CrawlMany eng leaders and platform engineers I speak with have already started on the “crawl” part of building a software factory by building simple automations using cloud agents. Think of these automations as Trigger → Agent Activity.For example:Issue reproduction and triage: have an agent look at all new issues that are filed and reproduce them and label them.Code review: Automatically review PRs as they are opened and leave commentsMonitoring: have an agent that responds to a Sentry alert and debugs and fixes an issueSelf-heal CI: fix broken CI by identifying PRs to roll back, merge conflicts to resolveAutoupdate docs: update user-facing documentation and generate changelogsVerification: browser-use and computer-use agents visually QA and verify changesSimple bug fixes: agents identify and fix simple user-reported…

{"@context":"https://schema.org","@type":"Blog","@id":"https://www.warp.dev/blog/category/company#blog","name":"Company posts","url":"https://www.warp.dev/blog/category/company","description":"Read the latest Warp blog posts filed under Company.","publisher":{"@id":"https://www.warp.dev/#organization"},"blogPost":[{"@type":"BlogPosting","headline":"We are now factory engineers, not product engineers","url":"https://www.warp.dev/blog/we-are-now-factory-engineers-not-product-engineers","datePublished":"2026-06-18T12:00:00.000Z","description":"This is the memo I shared with the Warp team about what building Warp needs to look like. We will focus less on interactive coding and more own automating software factories, and work with other companies to help them do the same."},{"@type":"BlogPosting","headline":"How Rectangle Health Built an AI Teammate That Writes Its Own Code","url":"https://www.warp.dev/blog/rectangle-health-self-improving-ai-teammate","datePublished":"2026-06-12T14:19:00.000Z","description":"Rectangle Health used Oz to build a self-improving AI teammate that takes issues from triage through merged PR. The teammate, named Rex, currently ships 35K+ lines of code per week …

{"@context":"https://schema.org","@type":"Blog","@id":"https://www.warp.dev/blog/category/engineering#blog","name":"Engineering posts","url":"https://www.warp.dev/blog/category/engineering","description":"Read the latest Warp blog posts filed under Engineering.","publisher":{"@id":"https://www.warp.dev/#organization"},"blogPost":[{"@type":"BlogPosting","headline":"Adopting the software factory model: crawl, walk, run","url":"https://www.warp.dev/blog/adopting-the-software-factory-model-crawl-walk-run","datePublished":"2026-09-15T12:00:00.000Z","description":"A crawl, walk, run path for adopting the software factory model: start with point automations, stand up an end-to-end cloud development loop, then scale to a closed-loop factory stack."},{"@type":"BlogPosting","headline":"Closing the loop with self-improving cloud software factories","url":"https://www.warp.dev/blog/agent-self-improving-software-factories","datePublished":"2026-08-27T15:00:00.000Z","description":"It's time to apply a true engineering mindset to deploying coding agents. Set up a closed-loop system in the cloud where every agent is tracked and measured against your own data and workflows, so you can tune your set…

{"@context":"https://schema.org","@type":"Blog","@id":"https://www.warp.dev/blog/category/product#blog","name":"Product posts","url":"https://www.warp.dev/blog/category/product","description":"Read the latest Warp blog posts filed under Product.","publisher":{"@id":"https://www.warp.dev/#organization"},"blogPost":[{"@type":"BlogPosting","headline":"The Factory Stack","url":"https://www.warp.dev/blog/the-factory-stack","datePublished":"2026-09-05T15:00:00.000Z","description":"Software factories should be built on an infrastructure stack that is open, composable and defined in code."},{"@type":"BlogPosting","headline":"Introducing Factory Benchmarks","url":"https://www.warp.dev/blog/warp-factory-benchmarks","datePublished":"2026-09-03T15:00:00.000Z","description":"Today we are launching Benchmarks in Warp Factories early access: agent benchmarks on your own code that drive direct improvements in software factory performance. Rather than guessing at what model and skill configurations are best, you can now easily test, measure, and put your findings into practice."},{"@type":"BlogPosting","headline":"Introducing Warp Factories - open, flexible infrastructure for building your software…

{"@context":"https://schema.org","@type":"CollectionPage","@id":"https://www.warp.dev/articles#collection","name":"Warp Articles","url":"https://www.warp.dev/articles","description":"Practical guides and reference material around Warp Factories, Software Factories, and cloud agent workflows.","publisher":{"@id":"https://www.warp.dev/#organization"},"mainEntity":{"@type":"ItemList","itemListElement":[{"@type":"ListItem","position":1,"item":{"@type":"TechArticle","headline":"How do you reduce AI coding agent costs without losing quality?","url":"https://www.warp.dev/articles/reduce-ai-coding-agent-costs","datePublished":"2026-09-14T16:00:00.000Z","description":"Measure cost per merged PR first, then pull four levers in order: route cheap task classes to cheaper models, trim the context each agent loads, cap retries and idle compute, and move verification earlier so failed work is caught before it burns more tokens. Routing pays the most, because most agent spend goes to frontier models doing work smaller models finish just as well."}},{"@type":"ListItem","position":2,"item":{"@type":"TechArticle","headline":"How do you avoid model and vendor lock-in with AI coding agents?","url":"htt…

Software FactoriesHow do you move from coding agents to a software factory?August 25, 2026You move from coding agents to a software factory by lifting agent work off individual laptops into governed cloud environments, then wrapping that work in a loop that triages, specs, implements, reviews and verifies it. Migrate one bounded workflow at a time, carry over the skills and harness choices your team already trusts, and keep humans at the decision points.Why individual agent productivity stops compoundingMost engineering organizations already made the first transition: developers write code with agents rather than by hand. That change raised individual output, but it left the gains stranded. The agent runs on one laptop, with one engineer's prompts, one engineer's tool configuration, and access to every system that engineer is logged into. When the session ends, the reasoning, the cost, and the quality signal are all gone.That is why the second transition looks less like adopting a better agent and more like building infrastructure. A software factory is an automation loop around the software development lifecycle — triage, spec, implement, review, verify, monitor — where cloud agen…

Model RoutingHow to use Claude Code with a Warp factoryAugust 27, 2026Can I use Claude Code with a Warp factory?Yes. Claude Code connects to a Warp factory in two directions: it can be the harness that factory agents use to execute work in the cloud, and your local Claude Code can push work into a factory — and pull it back — through the Factory MCP. You keep Claude Code's behavior and models; the factory adds the runtime, orchestration, human approvals, and measurement around it.A factory doesn't replace your coding agent — it schedules itThe confusion behind this question is that Claude Code and a software factory look like competing products. They aren't. Claude Code is a harness: the program that holds a model, reads your repo, runs tools, and edits files. A factory is the automation loop around the SDLC — triage, spec, implement, review, verify, monitor — that decides which harness runs which piece of work, on what schedule, with which human checkpoints.In Warp Factories, every stage of that assembly line is a separately defined agent, and each agent gets its own harness, model, and context. Warp is natively multi-model and multi-harness: an agent can run Warp's own agent for …

Software FactoriesWhat's the better platform for software factories: Warp or Cursor Cloud Agents?August 19, 2026What's the better platform for software factories: Warp or Cursor Cloud Agents?Cursor Cloud Agents run coding tasks on isolated cloud VMs — kicked off from the IDE, Slack, GitHub, Linear, or a phone — and Cursor says more than 40% of its own team's pull requests now come from them. That's a strong autonomous coding agent, not a software factory. Warp Factories sits a layer above: a control plane that routes work through triage, spec, implementation, and review agents, and can call a Cursor-style cloud agent as one harness option rather than being the entire workflow.What Cursor Cloud Agents areEach Cursor Cloud Agent, what Cursor used to call a background agent, gets a dedicated VM with a cloned repo, dependencies, secrets, and network access, so it can write code, run tests, and open a pull request without a laptop staying open (Cursor's Cloud Agents documentation). They support MCP servers, repo-committed hooks, computer use, and artifacts — screenshots, videos, and logs attached to the pull request (Cloud agent capabilities).Cursor Automations layer scheduled and event…

Software FactoriesWhat Are the Best Software Factory Platforms in 2026?August 21, 2026The best software factory platforms in 2026 aren't one product category. They split into the control plane that orchestrates a full SDLC loop — triage, spec, implementation, review, verification — and the coding agents that do session-level work inside that loop, including GitHub Copilot, Cursor, Claude Code, Aider, and Continue.What a software factory platform actually automatesA software factory platform implements an automation loop around the software development lifecycle: triage, spec, implementation, review, verification, shipping, and monitoring. Cloud agents handle the repeatable work inside each stage — turning a report into a reproducible spec, writing the implementation, reviewing the diff, verifying the fix actually works — while humans stay in the loop at the decision points that need judgment: approving a spec before code gets written, reviewing a pull request before it merges, deciding whether a low-risk fix ships on its own or waits for review.Three things sit underneath that loop, whichever platform runs it. A governed cloud runtime executes agent work somewhere other than a deve…

Software FactoriesWhat Is Agentic AI Governance, and What Does a Governance Framework Look Like?August 28, 2026Agentic AI governance is the set of policies and infrastructure controls that decide what an AI coding agent can access, how its actions are tracked, and who is accountable for its output. A working governance framework covers access scope, an audit trail, model and harness policy, data handling, and ongoing cost and quality accountability — enforced in infrastructure, not just written in a policy document.Why governance became urgentThe default way teams adopted coding agents was one developer, one bespoke agent, installed and configured on a laptop with access to everything that developer is logged into. That setup works until it doesn't: every agent is a separate security surface, there's no standard set of skills or MCPs across the team, and all the data the agents produce — conversations, decisions, the record of what an agent actually did — disappears the moment the session ends. Warp has called this out directly in its own account of why cloud software factories exist, describing it as a governance nightmare that scales faster than any control layer around it.The fi…

How do you choose which AI model to use for each coding task?September 14, 2026Choose per task class, not per organization. Group work into classes — triage, simple frontend changes, backend refactors, code review — then benchmark two or three candidate models on real tasks from your own repositories and route each class to the winner. A single default model overpays on easy work and underperforms on hard work.Why one default model is the wrong unit of decisionMost teams pick a model the way they pick a database: once, centrally, and then they live with it. That made sense when the choice was between two frontier models with a wide capability gap. It doesn't now. Open-weight models have closed enough of that gap that a large share of everyday engineering work — dependency bumps, test scaffolding, small UI changes, triage decisions — completes just as reliably on a model that costs a fraction of the frontier price.The cost of the wrong default compounds quietly. Every trivial task routed to a frontier model is an overpayment you never see itemized, and every hard task routed to a cheap model produces a plausible-looking diff that fails review. Both failure modes are invisible withou…

How do you avoid model and vendor lock-in with AI coding agents?September 14, 2026Treat lock-in as a per-layer question, not a single vendor decision. Your agent stack has distinct layers — factory definition, data and context, compute, inference, improvement, orchestration, access — and each carries its own switching cost. Keep ownership of the layers that are expensive to move (data, compute, inference) and accept managed services where exit is cheap.Lock-in is a layer-by-layer question"Are we locked in?" is unanswerable at the vendor level because a vendor is never one thing. A platform might hold your orchestration and nothing else, or it might quietly hold your agent traces, your skills, your compute, and your only path to inference. Those are very different exposures, and a single yes/no verdict hides the difference.The useful move is to decompose the stack and ask the question once per layer. Warp's factory stack breakdown lays out seven layers and argues they should be open (works with any model, agent, and hosting configuration), composable (adopt part or all), and defined in code (versioned, testable, revertible). Those three properties are also a good working definition …

How do you reduce AI coding agent costs without losing quality?September 14, 2026Measure cost per merged PR first, then pull four levers in order: route cheap task classes to cheaper models, trim the context each agent loads, cap retries and idle compute, and move verification earlier so failed work is caught before it burns more tokens. Routing pays the most, because most agent spend goes to frontier models doing work smaller models finish just as well.Measure cost per merged PR before you cut anythingToken spend is not a cost metric. It tells you what you consumed, not what you got, and teams that optimize it directly end up making agents cheaper and less useful at the same time. The number that survives scrutiny is cost per merged PR: total agent spend for a run, divided by work that actually shipped.Cost per merged PR has the useful property of punishing false economy. A cheaper model that produces diffs nobody merges raises the number rather than lowering it. Pair it with two companions — human intervention rate and defect or rollback rate — and you have enough to tell a real saving from a quality regression wearing a saving's clothes.You need per-run cost, model, task type, a…

Software FactoriesWhat Is an Agentic Development Environment (ADE)?August 31, 2026An agentic development environment (ADE) is a development tool — typically a terminal, IDE, or CLI — built around AI coding agents as first-class participants rather than as an add-on. In an ADE, an agent can read the codebase, run commands, edit files, and carry out multi-step tasks, while a developer supervises, steers, and takes over when needed, instead of only receiving inline code suggestions.Agentic vs. AI-assisted: the distinction that actually mattersMost developer tools added some form of AI years ago — autocomplete, a chat sidebar, an "explain this function" button. That makes a tool AI-assisted, not agentic. The dividing line is whether the AI can act on its own, across multiple steps, inside the same surface a developer already works in:AI-assisted: the model suggests text; the developer accepts, rejects, or edits it. Every action still originates from a human keystroke.Agentic: the model can plan a task, execute a sequence of commands or edits, observe the result, and adjust — with the developer approving, steering, or interrupting at checkpoints rather than authoring every step.An agent…

Software FactoriesWhat Tools Make Up a Modern Software Factory? (2026 Breakdown)August 11, 2026A modern software factory is made up of six tool layers: cloud runtimes and sandboxes, coding agents, orchestration, integrations, human-in-the-loop controls, and measurement, evals, and memory. Teams can assemble these components themselves, but platforms such as Warp provide a foundation for operating them as one governed workflow.A software factory turns the core development loop, which includes triage, specification, implementation, review, verification, shipping, and monitoring, into an automated system where agents perform work and developers provide judgment where it matters. This post will break down each of the layers that are required for a Software Factory.The six layers, at a glance:Cloud runtimes and sandboxes — cloud dev environments, Dockerized hosts, or Kubernetes running on AWS, GCS, Modal, or Daytona.Coding agents — Claude Code, Codex, Cursor, OpenCode, or Warp, connected to source forges like GitHub or GitLab via MCP.Orchestration and workflow engine — triggers agents from Slack, a ticket, or a schedule, then sequences triage, spec, implement, review, verify, ship, and …

SupportSupport & CommunityCopyCopy pageCopy page as Markdown for LLMsCopy agent promptCopy a prompt to guide you through this pageView as Markdown ↗View this page as plain textOpen in ChatGPT ↗Ask ChatGPT about this pageOpen in Claude ↗Ask Claude about this pageExport as PDFSave or print this pageCopied!# Support & Community ## Contact support ### Subscribers and Enterprise * **Technical issues or questions** - For bugs, credits, and other technical issues, email [[email protected]](mailto:[email protected]). * **Billing issues or questions** - For refunds, cancellations, and other billing questions, email [[email protected]](mailto:[email protected]). * **Enterprise** - Direct all feedback and issues to your designated Slack channel. For other ways to send feedback, gather logs, or file a bug report, see [Sending feedback and logs](/support-and-community/troubleshooting-and-support/sending-us-feedback/). ## Find your space ### Join the community Our Slack community is where workflows get shared, bugs get squashed, and ideas turn into features. Talk directly with Warp engineers and other developers. :::tip [**Join the Warp community on Slack**](https://go.warp.dev/join-preview) ::: Bui…

Terminal > Getting startedGetting started with WarpCopyCopy pageCopy page as Markdown for LLMsCopy agent promptCopy a prompt to guide you through this pageView as Markdown ↗View this page as plain textOpen in ChatGPT ↗Ask ChatGPT about this pageOpen in Claude ↗Ask Claude about this pageExport as PDFSave or print this pageCopied!# Getting started with Warp Warp is an [open source](https://github.com/warpdotdev/warp) **Agentic Development Environment** that combines a modern, high-performance terminal with powerful agents to help you build, test, deploy, and debug code. Agents in Warp are powered by the **Automation Platform**, which orchestrates agents locally or in the cloud at scale. <figure>  <figcaption>Warp and Warp Factories in the Agentic Development Environment.</figcaption> </figure> --- ## Warp Warp is where you work — a fast, modern terminal built for coding with agents. **Key capabilities:** * [**Terminal and Agent modes**](/agents/local-agents/interacting-with-agen…

ChangelogChangelogCopyCopy pageCopy page as Markdown for LLMsCopy agent promptCopy a prompt to guide you through this pageView as Markdown ↗View this page as plain textOpen in ChatGPT ↗Ask ChatGPT about this pageOpen in Claude ↗Ask Claude about this pageExport as PDFSave or print this pageCopied!# Changelog Submit bugs and feature requests on our [GitHub board!](https://github.com/warpdotdev/Warp/issues/new/choose) * [**2026**](/changelog/2026/) * [**2025**](/changelog/2025/) * [**2024**](/changelog/2024/) * [**2023**](/changelog/2023/) * [**2022**](/changelog/2022/) * [**2021**](/changelog/2021/)Summarize what's new and whether I need to take any action: https://docs.warp.dev/changelog/function e(){let e=document.getElementById(`copy-dropdown-wrapper`),t=document.getElementById(`copy-dropdown-trigger`),n=document.getElementById(`copy-dropdown-panel`),r=document.getElementById(`copy-toast`),i=document.getElementById(`copy-as-markdown`),a=document.getElementById(`copy-agent-prompt`),o=document.getElementById(`view-as-markdown`),s=document.getElementById(`export-as-pdf`),c=document.getElementById(`page-markdown-content`),l=document.getElementById(`agent-prompt-content`);if(!e||!t||!n…

EngineeringClosing the loop with self-improving cloud software factoriesAugust 27, 2026It's time to apply a true engineering mindset to deploying coding agents. There’s too much hand-waving around what agents are best, which models to use, and how to optimize ROI from coding agents over time. The solution is to set up a closed-loop system in the cloud where all of your agents are tracked and measured against your own data and workflows, so you can adjust your setup based on actual data and not vibes.The emerging category of infrastructure that supports this approach is the cloud software factory. Software factories are automation loops around the SDLC, comprised of agents that triage, spec, implement, verify, review, monitor, etc. Done right, these factories allow for true measurement, improvement and automation over time, and can help prove that you are doing agentic engineering the right way.Not all factory infrastructure is equal, though. As you evaluate options, you should look for the following:Factories should be defined as code, version-controlled, and have their definitions editable by agents.Factories must live in the cloud to support team access, central data storage, and…

ProductThe Factory StackZach Lloyd|September 5, 2026Software factories should be built on an infrastructure stack that is open, composable and defined in code.Open == works with any model, agent and hosting configurationComposable == you can adopt part or all of the stack, and use the pieces however you wantDefined in code == factory state is versioned, testable, and revertibleIn this article, I’ll lay out how we designed the stack for Warp Factories. The principles apply to anyone who is looking to move to a factory model.Let’s break this down layer by layer, starting with the foundation and working our way up:Factories-as-codeUnderlying the entire factory stack is a definition of the Factory-as-code, starting with factory.yaml, including:Agent config: Their prompts, skills, MCPs, model routing configsRunners: Information on what environments the factory runs in (e.g. how to launch factory agents)Repos: A set of repos corresponding to the product the factory operates onAutomations: Triggers and actions that drive agentic behavior, like scheduled crons, error monitor pings, taskboard status changes, GitHub pull request updates, etc.Integrations:What tools the factory integrates wit…

Software FactoriesHow Do Evals and Scorers Work in a Cloud Software Factory?September 1, 2026Evals and scorers are how a cloud software factory measures whether its coding agents are actually improving, rather than just producing output. A scorer is a defined rubric — built in or custom — that grades a sample of factory agent runs on dimensions like token cost, code quality, and defects caused. Evals run those scorers repeatedly across configurations so a team can compare, for example, one model or harness against another on the same task.Why Measurement Is a Distinct Layer from the Workflow ItselfA factory's triage-spec-implement-review-verify pipeline tells you that work moved through the system; it doesn't tell you whether the work was any good, or whether it's getting better or worse over time. Two factories can both report that 80% of PRs merged without human edits and be in very different states — one because agents are genuinely producing clean code, the other because reviewers have stopped catching real problems. Evals and scorers exist to answer the question a workflow can't answer on its own: is the factory actually improving, and on what basis would you know?What a Score…

Software FactoriesWhat's the better platform for software factories: Warp or Claude Code?August 19, 2026What's the better platform for software factories: Warp or Claude Code?Claude Code is Anthropic's coding agent — a CLI, desktop app, and set of cloud sessions that write and ship code, tied to Claude models. Warp Factories is the control plane that runs a governed triage, spec, implement, and review pipeline across any model or harness, including Claude Code itself. Teams building a factory need the latter; Claude Code is a component you run inside it.What is Claude Code?Claude Code has grown from a terminal agent into a real platform: a redesigned desktop app with a "Mission Control" sidebar for parallel sessions, Remote Sessions and Scheduled Tasks that keep running after a laptop closes, and "Routines" triggered by cron, webhook, or API, per VentureBeat's April 2026 test of the redesigned app.In August 2026, Anthropic made "auto mode" — letting Claude decide which actions are safe to take without asking permission — the default for Pro, Max, and Team plans (Anthropic, Aug 7 2026; TechCrunch, Aug 9 2026).Anthropic has said Claude Code's run-rate revenue surpassed $2.5 billion, …

Agent OrchestrationWhich terminal should you use for AI coding agents in 2026?September 4, 2026Pick by layer, not by brand. Ghostty, iTerm2, Kitty, WezTerm and Alacritty are terminal emulators that render an agent's interface well. Herdr is a runtime that keeps many agents alive and tells you which one is blocked. Warp is an agentic development environment that also hands local agent work to a governed cloud factory.What "agent terminal IDE" means in 2026Until recently a terminal was judged on input latency, escape-sequence correctness and font rendering. Then CLI coding agents — Claude Code, Codex, Gemini CLI, opencode, Cursor's CLI — became the main occupant of the pane, and three new questions started deciding the choice:Does it render an agent correctly? Full-screen agent TUIs need multi-line input (Shift+Enter), desktop notifications, clean Unicode, and ideally scrollback while an app owns the alternate screen.Does the session outlive you? Agent runs now last hours. A closed lid or a dropped SSH connection should not kill them.Can you see which agent is stuck? One agent fits in one window. Six agents across three repos do not, and the expensive failure is an agent sitting bloc…