Skip to main content
Blame is an automated code review and codebase intelligence platform built on deterministic Tree-sitter AST syntax parsing, multi-tier relational dependency graphs, and cross-encoder reranking. It equips engineering teams with high-confidence pull request reviews, granular review customization, automated CI/CD branch gates, multi-repository chat, and self-updating living documentation.

​
1. GraphRAG Codebase Context

Blame never reviews files in isolation. When analyzing changes, it maps relationships across your entire codebase using Tree-sitter AST queries and a multi-tier PostgreSQL dependency graph.
  • Impact Analysis: Traces upstream imports and downstream callers across files to evaluate how signature or logic changes ripple through your system.
  • Contract & Validation Verification: Detects when modified endpoints or services fail to invoke standard permission guards, schema validators, or audit events used across sibling modules.
  • Syntactic AST Chunking: Parses code by functional boundaries (functions, methods, classes, types) rather than arbitrary line counts.
Learn more about Graph-Based Context →

​
2. Automated PR Reviews & Calibrated Confidence

Blame reviews GitHub pull requests automatically upon creation or commit synchronization, providing thorough assessments without rubber-stamping.
  • Calibrated Score (1 to 5): Every PR receives an overall score (from 5/5 Production Ready down to 1/5 Critical Failures) with an explicit technical justification detailing edge cases, test gaps, and stability risks.
  • 1-Click GitHub Suggestions: Inline comments provide exact, indented replacement code blocks that can be committed directly from the GitHub conversation tab.
  • Modular PR Summaries: Configure which sections appear in the PR summary comment, including executive overviews, architectural evaluations, issue tables, sequence diagrams, and comments outside the active diff.
Learn more about PR Reviews →
Group 4 6

​
3. Review Strictness & Customization

Teams maintain full control over the depth and tone of Blame’s code reviews through workspace and repository settings.
  • Strictness Levels:
    • Low: Focuses strictly on critical runtime failures, security vulnerabilities, and crash-inducing bugs.
    • Medium (Default): Evaluates edge cases, missing guards, and architectural consistency.
    • High: Exhaustive review covering performance trade-offs, defensive patterns, and missing test coverage.
  • Custom Comment Header: Prepend custom messages or badges to review comments.
  • Draft PR Controls: Choose whether to run reviews automatically on draft pull requests or wait until they are marked ready for review.
Group 4 5

​
4. CI/CD Quality Gates & Status Checks

Prevent unstable or risky code from entering your primary branches by integrating Blame into GitHub branch protection rules.
  • GitHub Check Runs: Automatically creates check runs on GitHub commits reflecting the review outcome.
  • Required Confidence Threshold: Enforce minimum confidence scores (e.g. require at least 3/5 or 4/5) before the check passes.
  • Automated PR Status Comments: Post pass/fail review status summaries directly to PR discussion threads.
Group 4 2

​
5. Auto-Approve Low-Risk Pull Requests

Accelerate routine workflows by allowing Blame to automatically approve low-risk pull requests.
  • Risk-Capped Approvals: Set maximum allowable risk thresholds (Low, Medium, High, Critical) for auto-approvals.
  • Custom Approval Guidelines: Provide specific natural-language rules defining what qualifies for automated approval (e.g., dependency version bumps, documentation updates, or minor UI tweaks).
Group 4 3

​
6. Fix with AI Coding Agents

Bridge the gap between review feedback and bug remediation by dispatching Blame’s findings directly to your development environment.
  • Agent-Ready Fix Prompts: Every inline review comment includes an agentPrompt with file coordinates, line numbers, and concise fix instructions.
  • Integrated Agent Support: Format findings for tools like Cursor, Windsurf, Claude Code, Conductor, and Devin.
  • Direct GitHub Branch Fixes: Use Blame’s web chat agent to create branches and commit suggested changes directly back to your repository.
Group 4 1

​
7. Repository Memory & Auto-Discovered Rules

Blame adapts to your team’s conventions without requiring manual configuration.
  • Zero-Config Rule Discovery: Automatically scans and respects .cursorrules, .cursor/rules/*.mdc, CLAUDE.md, AGENTS.md, .coderabbit.yaml, blame.yaml, and .github/copilot-instructions.md.
  • Dashboard Memory Store: Store custom team rules with glob path targeting (e.g., services/auth/**) that apply to both automated reviews and chat sessions.
Learn more about Codebase Memory →
Group 1 4

​
8. Multi-Repository AI Chat

Inspect and debug complex systems spanning multiple codebases in a single conversational thread.
  • Cross-Repository Scope: Select multiple repositories simultaneously to trace API contracts between frontends, backend services, and shared libraries.
  • Voyage AI Reranking: Uses voyage-code-4 embeddings and rerank-3 cross-encoders to supply exact AST chunks into the conversation.
  • Direct GitHub Tooling: List PRs, inspect commit diffs, create branches, and commit fixes directly from chat.
Learn more about AI Chat →
Group 4 4

​
9. Team Workspaces & Notification Hub

Collaborate smoothly with built-in team governance and external alerting.
  • Role-Based Access Control: Manage permissions with Owner, Admin, and Member roles.
  • Slack, Linear & Discord Integrations: Receive review notifications, low-confidence alerts, and link PRs directly to Linear issue trackers.
  • Seat & Quota Management: Scale seats dynamically with transparent usage tracking.
Learn more Workspaces →
Group 6 1

​
Next Steps