The latest AI news and updates of Warp — Agentic development environment combining terminal, code editor, and AI agents for software development. Covering Warp's latest product updates, company news, and launches from the past 90 days.
Warp23h agoWarp adds agent debugging tools to Warp Factories, including a timeline for subagent handoffs with trace visibility, per-subagent token and compute cost breakdowns, and model-used-at-each-handoff tracking. A tool-by-tool flame graph for identifying wasted work is currently in internal testing. These panels provide granular visibility into agent actions, tool calls, and handoffs for inspecting individual runs.
WarpOct 21Warp open-sourced the merge-risk agent skill to address code review bottlenecks. The skill grades pull requests across five categories—alignment, impact, verification, rollback cost, and error detection—to identify low-risk changes. It installs via npx to automate risk assessment and help developers trust agent-generated code.
WarpOct 20Warp reports that three months of internal software factory adoption resulted in 40% more merged pull requests and a 67% reduction in cost per PR. The platform, which uses agent quality measurement, model benchmarking, and self-improvement loops, also achieved a 70% lower PR revert rate. Teams can now request early access to Warp Factories for their own development workflows.
WarpOct 15Warp is deploying proactive agent loops to create a closed-loop software factory. Agents now automatically investigate server errors from Sentry or Grafana to propose fixes, score past conversation traces to improve their own skills, and analyze customer interview transcripts via MCP to suggest roadmap features. These automations shift agents from reactive execution to proactive development and task selection.
WarpOct 14Warp defines a software factory as the centralization of a team's agent system into a single source of truth. This framework unifies repository access, secret management, sandbox environments, and model selections. By treating agent setups like CI or lint rules, this approach shifts the engineering role toward factory engineering, focusing on improving agent performance rather than manual prompting.
WarpOct 8Warp uses a multi-agent software factory to build its iOS app, orchestrating a foreman, an implementer on a MacOS cloud machine, and a code reviewer on Linux. The system runs agents in isolated Docker containers, using Slack for task intake and iOS simulator screenshots for verification. Many changes ship without requiring local Xcode builds or manual environment setup.
WarpOct 7Warp released Skill Doctor, an open-source skill that automates agent self-improvement by reviewing past coding conversations. The tool aggregates transcripts, scores them against rubrics for code quality and efficiency, and proposes merge-ready improvements to skill files. It supports Claude Code, Codex, and Warp agents, helping teams transition from manual coding to automated software factory workflows.
WarpOct 4Warp is detailing its internal multi-agent software factory, which orchestrates triage, implementation, verification, and code review agents to automate development. The system uses configuration-as-code to manage repositories and cloud access. Warp is currently benchmarking whether to maintain these specialized agent roles or consolidate the workflow into a single high-intelligence model.
WarpSep 30Warp released the Warp Factories configuration spec, a code-based repository for managing agent orchestration, automations, environments, and performance benchmarks. This spec allows engineers and agents to propose factory changes as pull requests, enabling automated testing and iterative improvements. Warp provides interactive guides and templates, with $10,000 in usage credits available for qualified organizations deploying the new configuration system.
WarpSep 30Warp integrated OpenAI's GPT-6.1 Sol into its terminal and agentic coding interface. The model appears in the picker, which groups options by speed and cost. Existing Codex or ChatGPT subscriptions fund agent conversations, and the agent confirms the model version during active sessions.
WarpSep 29Warp now supports signing in with a ChatGPT account directly within the Warp Terminal and Warp Agent CLI. This integration applies a subscription's included Work and Codex usage to agent conversations. Existing OpenAI subscription quotas now fund autonomous development tasks within the Warp environment.
WarpSep 29Warp has integrated Anthropic's Claude Sonnet 5.5 into the Warp terminal and Warp Agent CLI. The model is now selectable within the agentic coding interface for multi-step development tasks.
WarpSep 23Warp adds xAI's Grok 4.7 model to the Warp terminal and Warp Agent CLI. The agent performs multi-step coding tasks, including file creation and terminal command execution, using the model. Access requires connecting an existing Grok subscription to the Warp environment.
WarpSep 23Warp has integrated Anthropic's Claude Opus 5.5 into the Warp Terminal and Warp Agent CLI. The model is now available for selection within the agentic coding interface, supporting autonomous development workflows.
WarpSep 23Warp integrates OpenAI's GPT-6 Sol and GPT-6 Luna models into the Warp Terminal and Warp Agent CLI. The update adds these models to the existing model picker, providing access to the latest OpenAI reasoning and coding capabilities for terminal-based agentic workflows.
WarpSep 23Warp details a two-stage agent self-improvement loop where scorer agents grade past conversations on metrics like code quality and efficiency. Self-improvement agents then review failing runs, using the scorer’s report to propose PRs that update skills or AGENTS.md files. The /skill-doctor template provides a ready-made implementation compatible with harnesses like Claude Code, Codex, and Grok.
WarpSep 11Warp now supports the Grok Build CLI natively within its terminal. The integration adds rich input for agent prompts, including multi-cursor support and longer pastes. It also enables the /remote-control command for sharing agent sessions across devices, alongside direct access to file explorer and code review panels for managing agentic coding tasks.
WarpSep 9Warp reduced its internal cost-per-PR from $80 to $30 by switching its default agent model to OpenAI's GPT 5.6 Sol. Warp's internal benchmark, which replays real team tasks, found GPT 5.6 Sol delivered higher code quality than Claude Opus 5 at a 66% lower cost. The company is now offering early access to its benchmarking infrastructure.
WarpSep 3Warp launched Factory Benchmarks, a new feature in Warp Factories that generates evaluation suites from a team's own past agent runs. This allows teams to test agents against real workflows, reducing Warp's internal cost-per-PR by over 63%, from $80 to $30. The feature is now in early access, with $10,000 in usage credits available for select organizations.
WarpAug 30Warp released /skill-doctor v1, an open-source skill that crawls past Claude Code, Codex, and Warp agent conversations. It scores interactions against rubrics for efficiency and code quality, then proposes merge-ready diffs to improve agent skills. This tool automates the self-improvement loop for agent setups, with source code available on GitHub.
WarpAug 27Warp introduced self-improvement loops that allow agents to refine their own Skills by reviewing past conversations. The process scores interactions against defined criteria, isolates failures, and generates skill improvements. Warp Factories now supports this automated optimization, with early access available for organizations to receive up to $10,000 in usage credits.
WarpAug 22Warp launched Warp Factories, an open infrastructure for building cloud-based software factories. The platform supports any model or harness, defines factories as version-controlled code, and includes built-in memory and self-improvement agents. It provides cost and velocity metrics via API and SDK, with early access onboarding and $10,000 in usage credits available for select organizations.
WarpAug 13Warp integrated xAI’s Grok 4.6 model into its terminal and Agent CLI. The integration enables model access through the /connect-grok command using an active X Premium or SuperGrok subscription. This update brings the frontier model directly into the development environment for agentic coding tasks.
WarpAug 8Warp launched the /connect-grok command in its Agent CLI, enabling the linking of X Premium subscriptions. This integration provides direct access to Grok models within the terminal environment without consuming Warp’s own inference credits.
WarpAug 8Warp launched the Warp Agent CLI, a standalone terminal tool that reimagines agent orchestration. The interface makes subagents visible and allows configuration of models, harnesses, and environments using keyboard shortcuts. This CLI manages pty connections natively, enabling persistent sessions and remote agent workflows across any terminal environment.
WarpAug 6Warp launched custom model routers for its Agent CLI, which direct prompts to specific models based on plain-English rules. The system automatically routes tasks, such as sending planning requests to Claude or bug fixes to Kimi k3, according to configured logic. This feature supports frontier and open-weight models with at-cost inference pricing.
Frequently asked questions