AIBadger is a local context bridge for bringing codebase context to an AI chat. It scans a project on your machine, builds compact prompts, and helps apply AI-written file updates only after you review them.
AIBadger does not call AI providers, upload files, run telemetry, or require a network connection for normal use. You choose what to copy into your AI chat.
The first public release is planned as v0.1.0. Until packaged binaries are published, build from source:
go build -o badger ./cmd/badgerRelease-mode build:
go build -tags aibadger_release -ldflags="-s -w" -o badger ./cmd/badgerCheck the version:
./badger --versionExpected output:
badger v0.1.0
Run AIBadger from the root of the project you want to inspect:
badgerThe interactive workflow is:
- Map: AIBadger scans your project and prepares a topology prompt.
- Extract: You paste the AI's requested file commands back into AIBadger, and AIBadger prepares selected source context.
- Apply: You paste the AI's final response, review the write plan, and confirm before files are written.
Prompt 1 contains project structure and file paths, not source code. Prompt 2 contains selected source context based on the extraction commands you approve.
- All scanning and extraction runs locally.
- No telemetry is collected.
- No cloud sync is used.
- No source code is copied until you approve the handoff.
- No file writes happen until you review the preview and confirm.
- Secret and credential locations such as
.envand.gitare excluded from scanning.
AIBadger includes first-class scanning for:
- Go
- Java
- JavaScript
- TypeScript
- Python
It also includes generic project scanning for common source and configuration files when a first-class detector does not apply.
- AIBadger is a context bridge, not an AI provider or autonomous coding agent.
- It does not solve tasks by itself; your AI chat remains the reasoning layer.
- Extraction commands are intentionally simple:
FILE:,PREFIX:, andNEAR:. - Release builds hide development-only headless automation flags.
- Binary and generated files are intentionally excluded or minimized to keep prompts compact.
The public facade is:
github.com/PVRLabs/aibadger/pkg/badgerMost implementation packages remain under internal/ so the CLI and facade can evolve without exposing scanner, extractor, protocol, writer, or TUI internals as public API.