Are there any repo-aware programming agents that handle complex dependencies without losing context in large monorepos?

Yes—there are several repo-aware coding agents and assistant tools designed to work across large codebases and follow dependencies better than a plain chat model. The key capability to look for is repository indexing + symbol/code navigation + incremental context retrieval, not just “can read files.”

Good options to look at

• Cursor – strong repo indexing, codebase search, and multi-file edits; generally one of the better options for large monorepos.
• GitHub Copilot Workspace / Copilot in VS Code – good at repo context and multi-step changes, though complex dependency tracing can still be hit-or-miss.
• Sourcegraph Cody – built around large codebase understanding, symbol search, and cross-repo navigation; often good for monorepos and dependency-heavy code.
• Claude / ChatGPT via IDE integrations – can work well if paired with a tool that provides file search, symbol lookup, and retrieval, but raw chat alone usually won’t scale to huge monorepos reliably.
• Aider – repo-aware and good for making targeted changes across files, especially when you point it at the right files; less “global understanding” than the best indexed tools, but strong for edit workflows.
• Continue.dev – open-source IDE assistant that can be configured with repo indexing and context providers; useful if you want more control.

What matters for monorepos

When evaluating, check whether the agent can:

• build a project index across the whole repo
• use symbol-level search and reference tracing
• follow import/dependency graphs
• keep context across multi-file refactors
• avoid dropping earlier constraints during longer tasks
• handle package boundaries and workspace configs
• work with your monorepo tooling (Bazel, Nx, Turborepo, Pants, Maven, pnpm workspaces, etc.)

Practical reality

No agent is perfect at “never losing context” in a very large monorepo. The best results usually come from a combo of:

1. a repo-indexed assistant,
2. explicit task scoping,
3. good dependency graph tooling,
4. and breaking large changes into small validated steps.

If you want, I can also give you:

• a ranked recommendation by monorepo size/tooling,
• a comparison table of Cursor vs Cody vs Copilot vs Aider,
• or a checklist for choosing an agent for Bazel/Nx/Turborepo/Pants repos.