Sweep AI

Predicts the next code edit in real-time by indexing the entire project locally and analyzing surrounding context, delivering suggestions in milliseconds via a custom-trained 'Tab model' that learns from your codebase patterns. Suggestions are accepted by pressing Tab, enabling rapid code generation without context switching. Uses semantic understanding of code structure rather than simple pattern matching to generate contextually appropriate completions.

Enables searching across the entire indexed codebase using natural language queries (e.g., 'payment processing') rather than regex or keyword matching. Converts natural language intent into semantic embeddings and matches against indexed code embeddings to surface relevant functions, classes, and files. Results are ranked by semantic relevance and presented with code snippets and file locations.

Implements a flexible pricing model where autocomplete is unlimited on paid plans, but advanced features (code generation, chat, code review, web search) consume API credits. Free tier includes 1,000 autocompletes and $5 API credits; paid tiers ($10-60/month) include unlimited autocomplete and varying API credit allowances. Operates by tracking feature usage and deducting credits per request, with optional automatic top-up for continuous usage.

Analyzes differences between code branches (e.g., feature branch vs. main) and provides automated review feedback on changes, identifying potential issues, style violations, and improvement suggestions. Uses the indexed codebase context to understand how changes fit into the broader codebase patterns and conventions. Review results are presented inline in the IDE with actionable suggestions.

Enables browsing the web and fetching external content (documentation, API references, Stack Overflow answers) directly from the IDE without context switching. Queries are processed by Sweep's backend, results are fetched and summarized, and relevant information is presented inline. Supports OAuth 2.0/2.1 for authenticated access to restricted content. Added in v1.24 (Oct 6, 2025).

Resolves code definitions (function declarations, class definitions, variable assignments) across the indexed codebase and enables navigation to their source locations. Uses semantic understanding of code structure to identify definition locations even in complex inheritance hierarchies or dynamic code patterns. Integrates with IDE's 'Go to Definition' functionality to provide enhanced navigation with codebase context.

Connects to remote Model Context Protocol (MCP) servers to extend Sweep's capabilities with custom tools and data sources. Supports OAuth 2.0/2.1 authentication for secure access to remote servers. Enables integration with external services (databases, APIs, custom tools) without embedding credentials in the IDE. Added in v1.27 (Dec 1, 2025).

Indexes the entire project codebase locally, creating semantic embeddings and metadata for all code artifacts (functions, classes, files, patterns). Indexing is performed on-device in Privacy Mode or via cloud backend in standard mode. Indexed data enables fast semantic search, context-aware autocomplete, and code review without re-analyzing code on each operation. Index is updated incrementally as code changes.

Implements a credit-based billing system where different operations (chat, code generation, advanced completions, web search) consume API credits from the user's account. Credits are allocated based on subscription tier (Free: $5 initial, Basic: unspecified amount, Pro: more, Ultra: maximum). Consumption is tracked per operation and displayed in IDE. Provides usage alerts and limits enforcement to prevent overage.

Trains custom 'Tab model' LLMs on individual developer's codebase to learn project-specific coding patterns, conventions, and style. Training is performed using codebase data (with Privacy Mode ensuring data is never stored or used for general model training). Custom models are deployed locally or in cloud backend to provide personalized autocomplete suggestions that match the developer's coding style and project conventions.

Implements enterprise-grade security controls including SOC 2 Type II compliance certification, zero data retention policies, Privacy Mode for local-only operation, and encrypted communication with backend services. Designed for organizations with strict security and compliance requirements. Provides audit logs and compliance documentation for enterprise customers.