AIContextBuilder

What AIContextBuilder does

AIContextBuilder MCP server provides code intelligence for C# and .NET solutions. It analyzes solution files and source code, builds a semantic model, and answers questions about how symbols relate to one another. Supported analysis areas include callers, implementations, interface and override relationships, dependency-injection construction paths, tests, side effects, and the possible impact of changing a symbol.

The server can also assemble focused context for an AI coding task. Instead of returning an unrestricted source dump, it selects relevant symbols and their surrounding relationships, then renders the result as AI-Builder-MD: structured Markdown containing code, symbol relationships, architecture information, semantic metadata, and provenance.

Optional <ai> annotations in XML documentation let developers provide explicit information such as a type or method's role, layer, responsibility, stability, and side effects. These values supplement structural analysis and take precedence over heuristic inference where specified.

How it works

The processing path starts with a .sln, .slnx, or .slnf file and the associated C#, XAML, or AXAML documents. MSBuild and Roslyn semantic models produce facts about declarations, calls, type references, markup references, callers, implementations, and side-effect classifications. Tools then traverse or project this model for a particular question, or select a task-specific slice for context output.

A live MCP session keeps an analyzed graph and Roslyn workspace in memory for one server process. It sees saved files, not unsaved editor buffers. Source-only edits can use an incremental refresh; a full reload is available when needed or when explicitly forced. Separate server processes have separate live sessions.

Context-packing operations accept a token budget. Named seed symbols remain included while method detail and less-relevant surrounding material are reduced when space is limited. The output identifies structural omissions or caps, but the server does not certify that a remaining budget is sufficient to solve a task.

Tools and capabilities

The published MCP listing identifies these tools:

• aicb-csharp-context for C# and .NET code context.
• aicb-code-review for code-review workflows.
• aicb-code-simplifier for code-simplification workflows.
• aicb-usage-check for usage-check workflows.

The broader documented engine also includes operations for reading a symbol with dependencies and callees, exploring callers or implementations, packing context for a natural-language goal, preparing an edit with related tests and likely sibling symbols, and measuring the token cost of planned responses. The exact available tool set can vary by MCP profile; some opt-in codebase-memory tools are not exposed by default.

Limitations and notes

AIContextBuilder MCP server uses saved source files for live analysis and does not share one in-memory graph between separate processes. Remembered codebases can be persisted and recalled, but a recalled model has no live workspace, unreliable line numbers, and a reduced insight contract. Use a refresh when live precision is required.

The project is closed source. Its repository contains documentation, licensing information, and releases. The supplied material identifies the license as a custom EULA. The <Solution>.aicb.json configuration can be tracked with a solution and stores rules and choices, but not analysis results, sessions, or credentials.