
Turn GitHub issues into
repository-aware engineering plans.
Issue2Dev analyzes an issue with its repository context, then emits deterministic root-cause hypotheses, implementation plans, test strategy, and PRD artifacts locally, reviewably, and without GitHub writeback.
- Offline-friendly
- Deterministic
- Evidence-grounded
- No code writeback
- Read-only GitHub ingestion
- No provider key required
GitHub issues rarely contain enough context to implement safely.
Raw issue text creates ambiguity. While generic AI improvises single-prompt guesses, Issue2Dev deterministically compiles AST context, traces citations, and builds inspectable planning artifacts.
local/issue-42.json (ID: 42)
Deterministic Compilation
Traces execution paths, flags risk hotspots, and compiles diagnosis in real time.
Repository Ingestion
Build context automatically from file imports, AST definitions, and local dependencies.
Read-Only Sandbox
Zero code writeback. Pure read-only local analysis with untouched git working tree.
Offline Output Artifacts
Emits clean, structured markdown diagnostics and PRDs to .issue2dev/ output trees.
Grounding & Citations
Cites source file lines and AST locations for every single planning hypothesis generated.
How Issue2Dev processes issues
A pipeline structured for transparency. Every step exposes evidence paths.
Ingest Issue
Ingests raw issue input securely from local JSON configurations or directly via read-only GitHub API parameters.
Build RepositoryContext
Heuristically maps repository stacks, architecture signatures, risk zones, and component dependency graphs.
Analyze
Calculates impact surface, assigns severity metrics, estimates risk profiles, and scores planning priorities.
Diagnose
Emits localized root cause hypotheses, maps solution pathways, and drafts integration checkpoints.
Generate Artifacts
Compiles reviewable markdown summaries and clean, schema-aligned JSON configuration trees under .issue2dev/
Review
Allows developers to review complete project-scoped strategies side-by-side with original codebase branches.
Why Issue2Dev?
Review issues like code, before writing code.
Context You Won’t Fake Anywhere
Issue2Dev builds repository context first — labeling uncertainty, preserving provenance, and producing reviewable planning artifacts engineers can inspect like code, not a single-prompt guess dressed up as an answer.
Full AST & Dependency Mapping
Maps cross-file imports, symbol definitions, and hierarchy chains.
Provenance-Preserved Planning
Every plan references exact source files and deterministic line boundaries.
Zero Code Writeback
Strictly read-only sandbox with zero destructive file alterations.
How does it compare?
Evaluating issue analysis strategies across context, accuracy, and infrastructure.
| Capability | Raw Issue | Generic LLM Summary | Issue2Dev |
|---|---|---|---|
Repository-Aware Context Detects file structures, languages, frameworks, and import dependencies. | Repository-Aware Context - Raw Issue: Not supported / No | Repository-Aware Context - Generic LLM Summary: Not supported / No | Repository-Aware Context - Issue2Dev: Supported / Yes |
Deterministic & Reproducible Running it multiple times yields matching, evidence-linked results. | Deterministic & Reproducible - Raw Issue: Not supported / No | Deterministic & Reproducible - Generic LLM Summary: Not supported / No | Deterministic & Reproducible - Issue2Dev: Supported / Yes |
Evidence & Provenance Tracking Attributes facts to specific lines of code or workspace configurations. | Evidence & Provenance Tracking - Raw Issue: Not supported / No | Evidence & Provenance Tracking - Generic LLM Summary: Not supported / No | Evidence & Provenance Tracking - Issue2Dev: Supported / Yes |
Confidence-Labeled Heuristics Explicitly states prediction certainty scores (e.g. Affected Files). | Confidence-Labeled Heuristics - Raw Issue: Not supported / No | Confidence-Labeled Heuristics - Generic LLM Summary: Not supported / No | Confidence-Labeled Heuristics - Issue2Dev: Supported / Yes |
Offline-Friendly (No Provider Call) Does not require internet access, external SaaS platforms, or remote keys. | Offline-Friendly (No Provider Call) - Raw Issue: Supported / Yes | Offline-Friendly (No Provider Call) - Generic LLM Summary: Not supported / No | Offline-Friendly (No Provider Call) - Issue2Dev: Supported / Yes |
Treats Issue Text as Untrusted Filters claims, checks validation, and safeguards inputs before assembly. | Treats Issue Text as Untrusted - Raw Issue: Not supported / No | Treats Issue Text as Untrusted - Generic LLM Summary: Not supported / No | Treats Issue Text as Untrusted - Issue2Dev: Supported / Yes |
No Vendor/API Key Required Can be executed locally without subscribing to API platforms or models. | No Vendor/API Key Required - Raw Issue: Supported / Yes | No Vendor/API Key Required - Generic LLM Summary: Not supported / No | No Vendor/API Key Required - Issue2Dev: Supported / Yes |
Repository-Aware Context
Detects file structures, languages, frameworks, and import dependencies.
Deterministic & Reproducible
Running it multiple times yields matching, evidence-linked results.
Evidence & Provenance Tracking
Attributes facts to specific lines of code or workspace configurations.
Confidence-Labeled Heuristics
Explicitly states prediction certainty scores (e.g. Affected Files).
Offline-Friendly (No Provider Call)
Does not require internet access, external SaaS platforms, or remote keys.
Treats Issue Text as Untrusted
Filters claims, checks validation, and safeguards inputs before assembly.
No Vendor/API Key Required
Can be executed locally without subscribing to API platforms or models.
Try the CLI locally
Install and run the Issue2Dev analyze compiler offline or connected to GitHub.
CLI Constraints
- Output directories must live under a
.issue2dev/path (enforced). - The CLI currently exposes the
analyzecommand. GITHUB_TOKENis optional for public issues, but recommended to avoid rate limits.
Deterministic Local Output Bundles
Output planning artifacts
Issue2Dev compiles structured files containing diagnostics, plans, and requirements for direct peer review.
Reviewable engineering plans. Outputs are grouped in folders, allowing developers to inspect diagnostic details before writing any code.
Designed to be inspectable,
bounded, and honest.
Issue2Dev is built for developers who audit their tools. We never write back to your repository, never modify your source files, and never make silent SaaS provider calls.
Read-Only Ingestion
All interactions with GitHub APIs are read-only. No write tokens are requested or utilized in execution.
Isolated Local Writes
Writes planning files exclusively under the local .issue2dev/ workspace folder. Code files remain unmodified.
Sanitized Entanglements
Treats issue text as untrusted. Normalizes variables and redacts environment properties automatically.
Inspectable Provenance
Every generated PRD is complete with grounding citations pointing back to source files and heuristics.
Verification Hash: SHA-256 Verified Offline Executable
Zero Server Storage: No Cloud Backends
Limitations of v0.1
We are upfront about what Issue2Dev can and cannot do. This is a technical utility, not a magic fix.
v0.1 is intentionally small
Only the single 'analyze' command exists. There are no additional interactive loops or multi-command features in this version.
No AI/LLM integration yet
All Repository Intelligence is 100% heuristic-driven, AST-based, and static rule-grounded. No generative neural nets are active.
Predictions are heuristics
Affected-file predictions are educated recommendations with confidence weights. They are not compile-time absolute guarantees.
GitHub API dependency
Remote GitHub issue retrieval requires active network connections and conforms to GitHub's standard hourly rate limits.
Primary Markdown Outputs:analysis.mdprd.md
Development roadmap
Planned future directions. These capabilities are not yet active in v0.1.
3 planned
Distribution
- GitHub Action Integration
- GitHub CLI (gh) extension wrapper
- NPM-published global executable binary
4 planned
Intelligence
- Semantic search & codebase embeddings
- Deep import AST & dependency path resolver
- Multi-language architecture pattern scans
- Duplicate issue & regression detection
3 planned
Generation & Providers
- Optional OpenAI / Anthropic provider hooks
- Offline local models integration (Ollama / Llama.cpp)
- Additional artifact schemas (issue templates)
4 planned
Interfaces
- Model Context Protocol (MCP) server support
- Visual Workspace Dashboard UI
- VS Code & JetBrains extension modules
- Pipeline plugins / lifecycle hooks API
Review issues like code,
before writing code.
Start building repository context first. Build and review deterministic planning artifacts on your local terminal today.