OpenMate Onboarding

Find a contribution you can actually start.

OpenMate analyzes a public GitHub repository, combines what it learns with your skills, interests, experience, and available time, and recommends a real issue with a practical starting point.

Start analyzing a repository →

Overview

What OpenMate Does

A deterministic context pipeline combined with open-weight AI reasoning to ground your first contribution.

OpenMate bridges the gap between wanting to contribute to open source and actually submitting your first merged pull request. Here is the complete workflow:

Choose a public repository: You provide the GitHub URL of a repository you want to explore.
Describe your profile: You share your technologies, focus interests, experience level, and available hours.
Bounded context extraction: OpenMate inspects manifests, key entrypoints, documentation, and open issues within a strict character budget.
Architectural analysis: Google Gemma 3 27B analyzes the codebase structure, tech stack, and contribution conventions.
Deterministic issue ranking: OpenMate filters real open GitHub issues based on label signals, issue detail, and your profile match.
Actionable recommendation: You receive "Your First Contribution" with fit reasoning, verified entrypoint source files, and setup instructions.
Ask OpenMate: You ask follow-up questions about files, setup commands, or architecture through a repository-grounded assistant.

Real Issues Only

OpenMate does not invent fake or synthetic issues. Every recommendation is drawn directly from active, open issues fetched from the repository on GitHub.

Boundaries

What OpenMate Does Not Do

Setting clear, practical expectations about the scope of this tool.

To keep your experience predictable and your environment secure, OpenMate has clear functional boundaries. OpenMate does not:

Find arbitrary repositories for you: You bring the repository URL you want to explore. OpenMate is not a marketplace or general repository discovery engine.
Write, edit, or submit code: OpenMate guides you on where to begin, but writing code and creating the pull request remains your creative work.
Create pull requests or modify GitHub repositories: OpenMate is strictly read-only and requires zero write permissions to GitHub.
Execute repository code or run shell commands: Untrusted repository code is never run, compiled, or evaluated on OpenMate servers.
Browse arbitrary third-party websites: Network requests are strictly bounded to the public GitHub API and vector search endpoints.
Guarantee issue availability: Open issues can be claimed, assigned, or resolved by other community members at any time.

Guide

Step-by-Step Walkthrough

How to configure your onboarding inputs for the most relevant results.

Step 1: Choose a Public GitHub Repository

Provide a standard public GitHub repository URL in the format https://github.com/owner/repo.

Good Choices:

Projects you actively use or find interesting
Repositories with active maintenance and open issues
Codebases using technologies you know or want to practice

Avoid:

Private or internal repositories
Subpages like /issues/123 or /pulls
Non-GitHub URLs (GitLab, Bitbucket, self-hosted)

Step 2: Describe Your Contributor Profile

OpenMate uses four profile dimensions to match you with suitable issues:

1. Skills & Technologies

Declare the languages, frameworks, or tools you are comfortable reading or writing. You can rate each skill:

BeginnerBasic familiarity, learning syntax
IntermediateComfortable reading & writing idioms
AdvancedDeep architectural and debugging expertise

Note: OpenMate does not independently verify your skills. Be honest about your current level to get well-matched issues.

2. Contribution Focus Areas

Select the specific subsystems or categories of work that interest you:

FrontendUI, components, state, styling

BackendAPI routes, data processing, server logic

Full StackEnd-to-end integration and features

DocumentationGuides, API reference, onboarding examples

TestingUnit tests, mock boundaries, regression coverage

Developer ToolsCLI tools, build pipelines, DX improvements

PerformanceMemory optimization, latency reduction, caching

AccessibilityScreen readers, keyboard navigation, WCAG

DevOps & CIWorkflows, containerization, deployment configs

3. Available Time

Estimate how many hours you plan to spend on this contribution (1 to 40 hours). This helps OpenMate distinguish between quick starter fixes and more involved feature tasks. It is not an exact time prediction.

4. Open-Source Experience

Choose your familiarity with open-source workflows. This informs how conservatively OpenMate filters issue difficulty and architectural scope:

First-time contributorNew to open source; seeking small, well-scoped starter issues
I've contributed a few timesComfortable with Git forks, branching, PR workflows, and code reviews
Experienced contributorRegular open-source author or contributor ready for non-trivial refactors

Pipeline

What Happens During Analysis

Understanding OpenMate's bounded ingestion pipeline.

When you click "Analyze Repository", OpenMate runs a multi-stage context extraction pipeline:

Metadata & Default Branch: Queries repository basics (stars, description, primary language, default branch) via GitHub REST API.
Project File Tree: Ingests the top-level directory structure, filtering out test fixtures, lockfiles, and generated build artifacts.
Key Documentation: Extracts the README, CONTRIBUTING guides, and license files to understand local setup and community conventions.
Dependency Manifests: Inspects files like package.json, Cargo.toml, or go.mod to determine libraries, scripts, and runtime requirements.
Open GitHub Issues: Fetches open issues, prioritizes those with labels like good first issue, help wanted, or documentation, and extracts their description text.

Bounded Context Enforcement

OpenMate does not upload entire repositories blindly. It deterministically compiles an essential context snapshot capped at approximately 60,000 characters, prioritizing documentation, dependencies, and issue bodies.

Results

Your First Contribution Dashboard

What to expect when the analysis completes.

Once analysis finishes, you are navigated to /repo, where you receive a comprehensive onboarding plan:

Your Recommended Issue

The primary recommended GitHub issue with a concise summary of why it fits your skills, estimated scope category (small, medium, large), and direct link to GitHub.

First Files to Read

Specific entrypoint source files deterministically verified against the real repository tree so you know exactly which file to open first.

Local Setup & Prerequisites

Prerequisites, package managers, and verification commands extracted from repo manifests and documentation to get your local environment running.

Architecture & Concepts

Subsystem overview, tech stack breakdown, and a glossary of repository-specific terms to orient you before writing code.

Note: Issue titles, issue numbers, and URLs are validated directly against real GitHub issues. OpenMate personalizes the explanation and starting plan, but the issue itself is authentic.

Interactive Assistant

Ask OpenMate (Repository Assistant)

Context-grounded follow-up chat powered by Backboard RAG and Gemma 3 27B.

On the results page, you can open Ask OpenMate to ask repository-specific questions. Example questions contributors often ask:

> "Which files should I read before writing code for this issue?"

> "Where would I likely add a new test for this feature?"

> "How does the data flow between the API router and database layer?"

> "What commands should I run locally to verify my changes?"

> "Why did you choose this issue over others in the repo?"

How Ask OpenMate works behind the scenes:

The repository context snapshot is indexed into a temporary Backboard thread.
Responses are strictly grounded in the ingested files, README, and issue details.
Persistent cross-user memory is disabled to protect contributor privacy.
OpenMate cannot execute terminal commands or make changes to GitHub.

Architecture

How Recommendations are Chosen

Two-stage pipeline: deterministic pre-filtering followed by open-weight AI evaluation.

Stage 1: Deterministic Filtering

Heuristic Issue Scoring

Before any AI inference takes place, OpenMate evaluates all open issues in the repository against several deterministic criteria:

Beginner-friendly labels (good first issue, help wanted, up-for-grabs)
Presence of clear description bodies and reproduction steps
Skill keyword overlap with your declared technologies
Alignment with your chosen contribution focus areas
Scope fit against your declared available time and experience

This scores and narrows hundreds of issues down to a focused shortlist of at most 8 real candidate issues.

Stage 2: Gemma 3 27B Reasoning

Deep Architectural Matching

Google Gemma 3 27B evaluates the shortlisted candidates against your complete profile and the extracted repository context:

Selects the single highest-fit primary recommendation
Generates personalized explanation of why your background fits
Identifies the exact starting entrypoint source files to open first
Synthesizes concrete, bounded investigation steps

The AI response is validated through strict Zod schemas before being rendered.

Security

Security & Privacy

A defense-in-depth architecture designed for untrusted open-source data.

OpenMate treats all repository URLs, source blobs, and issue bodies as untrusted input. Key architectural guardrails include:

Server-Side Credentials: All GitHub and Backboard API tokens stay server-side. Zero API keys are ever sent to client browsers.
SSRF Defense: Repository URLs are restricted to valid GitHub repositories. Internal, private, loopback, and metadata IP addresses are strictly blocked.
No Code Execution: Repository code is strictly treated as text for analysis. OpenMate never compiles, runs, or executes repository files.
Zod Output Validation: All AI completions are parsed and validated against strict schemas before rendering. Hallucinated file paths are checked against the repository tree.
Transient Chat Memory: Follow-up chat threads are scoped only to your active browser session. Persistent cross-user memory is turned off.
Client Storage Safety: Session data is saved only in your browser's sessionStorage, which is isolated to your tab and automatically expires after 24 hours.

For complete technical specifications, see our open-source codebase on GitHub (Ismailco/OpenMate).

Practical Guide

Limitations & Tips for Better Results

Practical advice to get the most value out of OpenMate.

Current Limitations

Public GitHub only: GitLab, Bitbucket, and private repositories are not currently supported.
Requires open issues: If a repository has zero open issues, OpenMate cannot recommend a contribution.
Context bounds: In massive monorepos, the 60,000-character context prioritizes root and main packages; deeper subpackages may be summarized.
Analysis latency: Multi-stage deterministic ingestion and Gemma 27B reasoning typically take 30 to 50 seconds to complete.
Unclaimed issues: OpenMate cannot reserve issues or guarantee that someone else won't submit a PR before you.

Tips for Better Results

Pick projects you use: You will understand requirements much faster if you have used the tool before.
Be honest about skills: Listing intermediate for skills you just started learning can lead to overly complex recommendations.
Focus your interests: Picking 1 to 3 specific focus areas yields more targeted matches than selecting all 9.
Read the GitHub issue thread: Check recent comments on GitHub to verify maintainers still welcome a fix before writing code.
Follow CONTRIBUTING.md: Respect project conventions for commit messages, testing requirements, and PR templates.

Ready to find your starting point?

Paste any public GitHub repository and tell OpenMate your skills to receive your personalized onboarding plan.

Find my first contribution →Back to OpenMate

Loading OpenMate workspace...