ApexCode AI Assistant
Bring your own API key. Review the plan. Investigate bugs inside Visual Studio.
ApexCode brings AI chat, planning, coding, and debugging into your IDE, with support for 7 LLM providers and no subscription fee. This update improves Plan Mode reliability, introduces a dedicated plan document for reviewing proposed changes, and adds Debug Mode for investigating failures with live debugger evidence.
What’s New
- Plan Mode bug fixes. Improved reliability across the planning, review, and approval workflow.
- Dedicated plan document. Review the proposed changes, affected files, decisions, risks, and open questions in a dedicated view. Add comments to individual plan items, provide overall feedback, and request revisions before proceeding. Browse revision history, compare changes, and copy or export the plan as Markdown.
- Debug Mode. Investigate failures using Visual Studio’s debugger. ApexCode can set breakpoints, step through execution, inspect call stacks and supported local or argument values, and use breakpoint evidence to guide the investigation. Supported reproduction workflows provide an editable approval preview before execution.
Why ApexCode
- No subscription. Bring your own key. You pay your provider’s usage charges — no token markup or platform fee. Connect OpenAI, Azure OpenAI, OpenRouter, DeepSeek, Gemini, NVIDIA, or Vercel.
- Plan before it acts. Plan Mode explores your workspace read-only and presents a dedicated implementation plan for review. Comment, request changes, and approve the plan before implementation begins. Read-only restrictions are enforced through tool access.
- Debug with runtime evidence. Use Debug Mode to investigate reported failures through source inspection, breakpoints, stepping, and live debugger information. Debugger tools follow your configured permissions, and reproduction requests require explicit approval.
- Built for real codebases. Large files are read in paginated line ranges and cached by last-write-time. Older conversation history is automatically evicted to stay within your provider’s context limit.
- Talks to your database, safely. Connect to SQL Server to inspect schemas and run validated, read-only SELECT/CTE queries. Writes and stored procedures are blocked, and results are capped at 100 rows.
- Deep IDE awareness. Uses Roslyn to read diagnostics, validate builds, and apply code fixes directly to your syntax tree.
- Checkpoints, not regret. Automatic workspace snapshots before tool-using AI turns let you review diffs and revert code changes.
- Custom personas. Define roles such as “Angular Developer” or “SQL Reviewer” to tailor the assistant’s behavior to your project or task.
Core Features
| Feature |
What it does |
| Multi-provider chat (BYOK) |
Connect any of 7 supported LLM providers with your own API key |
| Plan / Act / Debug modes |
Explore and plan, implement approved work, or investigate failures |
| Dedicated plan document |
Review proposed changes, comment, request revisions, and approve implementation |
| Plan revision history |
Browse earlier revisions, compare changes, and export Markdown |
| Live debugger integration |
Set breakpoints, step through code, and inspect supported runtime evidence |
| Approved debug reproduction |
Review and edit supported HTTP requests or method harnesses before execution |
| Paginated file reading |
Handle large files with cached, bounded reads |
| SQL Server integration |
Inspect schemas and run validated read-only queries, capped at 100 rows |
| Roslyn-powered fixes |
Read diagnostics, validate compilation, and apply syntax-aware fixes |
| Checkpoints & diffs |
Review and revert AI-driven workspace changes |
| Custom personas |
Tailor the assistant’s behavior per project or task |
| Token budget management |
Automatically trim older history to fit provider context limits |
Requirements
- Visual Studio 2022 (or later, where supported)
- An API key from a supported LLM provider: OpenAI, Azure OpenAI, OpenRouter, DeepSeek, Gemini, NVIDIA, or Vercel
- SQL Server connectivity is optional and required only for database features
- Live debugging requires a project that Visual Studio can debug; reproduction support depends on the project, runtime, and application setup
| |
