Warp Terminal

Converts natural language descriptions into executable shell commands using frontier LLM models (OpenAI, Anthropic, Google) with codebase context awareness. The system indexes the user's codebase to understand project structure, environment variables, and installed tools, then generates contextually appropriate commands that account for the specific development environment rather than generic suggestions. Execution happens directly in the terminal with user review before running.

Provides real-time command completion suggestions as users type, with syntax highlighting and contextual awareness of available commands, flags, and file paths in the current directory. The autocomplete engine understands shell syntax and integrates with the system's available commands and environment, displaying rich formatting that makes complex commands easier to construct. Completions are ranked by relevance based on usage history and context.

Implements configurable data retention policies where users can enable Zero Data Retention to prevent Warp from storing conversation history, command logs, or AI interaction data. Free tier allows individual configuration of Zero Data Retention, while Business tier enforces team-wide Zero Data Retention automatically. Data retention settings apply to cloud conversation storage and cloud agent execution logs.

Implements a usage-based credit system where AI features consume credits based on LLM API calls and cloud agent execution. Free tier includes limited free AI credits, Build tier provides 1,500 credits/month, and Max tier provides 12x credits (18,000 credits/month implied). Credits can be reloaded with volume-based discounts on Build tier and above. The credit-to-token conversion rate and per-feature credit costs are not documented.

Supports team collaboration with Business tier capped at up to 50 seats, enabling multiple team members to share sessions, collaborate on code review, and access shared cloud agents. Team-wide settings like Zero Data Retention enforcement and shared codebase indexing are available on Business tier. Seat-based licensing enables cost control for team deployments.

Enables interactive, multi-step task execution where an AI agent (Claude Code, Codex, OpenCode, or custom agents) can plan, execute commands, review results, and iterate based on feedback. Users can steer the agent mid-task, approve or reject proposed actions before execution, and maintain a conversation history across multiple turns. The system tracks all runs as auditable, shareable sessions stored in Warp Drive with full context preservation.

Generates and refactors code across a user's codebase using indexed project context, including file structure, dependencies, coding patterns, and environment configuration. The system understands the codebase structure through indexing (limits vary by tier) and can propose changes that align with existing patterns and conventions. Built-in code editor with LSP (Language Server Protocol) support, syntax highlighting, and file tree navigation enables inline code review and modification.

Provides an interactive code review experience where AI can analyze proposed changes, suggest improvements, and explain reasoning. The code review panel integrates with the terminal's block-based interface, displaying diffs alongside AI commentary and allowing reviewers to approve, request changes, or steer the AI mid-review. Reviews are tracked as part of shareable sessions in Warp Drive.

Enables scheduling of cloud agents to run on recurring schedules or trigger via external webhooks (Slack, Linear, GitHub, custom webhooks). Agents execute in the cloud with full audit trails, and results are tracked and shareable. The system supports concurrent agent execution and integrates with external platforms for event-driven automation without requiring local infrastructure.

Enables users to share complete terminal sessions (including command history, output, and AI agent reasoning) with team members via Warp Drive, a cloud-based collaboration platform. Shared sessions preserve full context and are accessible to team members with appropriate permissions, enabling asynchronous collaboration on debugging, code review, and task execution. Sessions are persistent, searchable, and can be referenced across projects.

Automatically reads and understands environment variables, Git configuration, and project metadata to provide context-aware suggestions and command generation. The system can detect the current Git branch, worktree, and repository state, and uses environment variables to tailor command suggestions to the specific development environment. This context is integrated into codebase indexing and command generation.

Replaces traditional line-based terminal output with a block-based interface where each command and its output are visually separated and independently navigable. Blocks can be collapsed, expanded, searched, and referenced, making it easier to navigate long terminal sessions and find specific outputs. The interface integrates with AI features, displaying command generation, code review, and agent reasoning alongside execution results.

Allows users to provide their own API keys for OpenAI, Anthropic, and Google LLMs, enabling cost control and model selection flexibility. Build tier and above support BYOK (Bring Your Own Key), while Enterprise tier supports BYOLM (Bring Your Own LLM) for self-hosted model deployment. Users can configure different models for different tasks and monitor their own API usage and costs.