GitHub Research is an open-source tool for researching GitHub trending repositories, newly emerging projects, and keyword-driven repository activity. It uses Decodo's Web Scraping API to retrieve public GitHub pages and your chosen LLM provider to turn normalized repository data into structured research reports.
Unlike leaderboard-style trending trackers, GitHub Research can start from a keyword or concept and return recurring themes, notable repositories, and historical changes instead of only a ranked list.
GitHub Research is not affiliated with or endorsed by GitHub, Inc.
Choose Trending, Emerging, or Keyword research from the main screen.
GitHub Research turns public repository activity into a structured research report instead of stopping at a leaderboard.
- Three research modes. Explore GitHub Trending, discover newly created repositories that meet traction filters, or start from a keyword or concept.
- LLM-synthesized findings. Get a research summary, recurring themes, notable repositories, and deterministic repository metrics in one report.
- Longitudinal monitoring. Save any research configuration as a recurring monitor and compare later snapshots with earlier runs.
- Verifiable outputs. Keep GitHub source links and repository-level data visible while deterministic metrics remain separate from LLM interpretation.
- Portable reports. Export completed research as Markdown or JSON, with every run stored in History for later review.
Example Emerging report generated on September 18, 2026.
- Choose a research mode. Select Trending, Emerging, or Keyword and configure the relevant filters.
- Build the GitHub research request. The backend converts the configuration into GitHub Trending or repository-search URLs.
- Retrieve public GitHub pages. Decodo Web Scraping API retrieves the page HTML used for the research run.
- Normalize repository data. Deterministic parsers extract repository metadata into a shared repository-snapshot format.
- Generate the research report. Your selected LLM provider receives the normalized current dataset, deterministic metrics, and relevant previous snapshots to produce themes and notable findings.
- Store the run. MongoDB stores the configuration, repository snapshots, and generated report for History and later comparisons.
- Review or export the result. Open the report in the UI or download it as Markdown or JSON.
The current research flow does not use the GitHub REST API as its data source. It retrieves public GitHub pages through Decodo Web Scraping API.
Before installing the project, make sure you have:
- Bun 1.2.5 or newer
- Docker or another Docker-compatible runtime
- A Decodo Web Scraping API Basic Auth token – create a Decodo account
- At least one supported LLM provider API key:
- Anthropic
- OpenAI
- Google Gemini
If you've just installed Bun, open a new terminal window or reload your shell configuration before running
bun install.Examples:
- zsh:
source ~/.zshrc- bash:
source ~/.bashrc
git clone https://github.com/Decodo/github-research.git
cd github-researchbun installCreate a local .env file from the included example:
cp .env.example .envFiles beginning with a dot may be hidden by default. If you don't see .env in the project folder:
- On macOS Finder, press Cmd+Shift+.
- On most Linux file managers, press Ctrl+H
- On Windows, use View → Show → Hidden items if needed
Open .env and add your Decodo token, select an LLM provider, and add the API key for that provider.
# Backend
PORT=5002
PUBLIC_API_BASE_URL=http://localhost:5002
PUBLIC_FRONTEND_URL=http://localhost:5274
# Database
MONGO_PORT=27018
MONGODB_URI=mongodb://localhost:27018/github-research
# Decodo Web Scraping API
# Paste the Basic Auth token only, without the word "Basic".
DECODO_AUTH_TOKEN=
DECODO_SCRAPER_ENDPOINT=https://scraper-api.decodo.com/v2/scrape
# LLM provider: claude | openai | gemini
LLM_PROVIDER=claude
LLM_MODEL=
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
GEMINI_API_KEY=Only the API key for the selected LLM_PROVIDER is required. LLM_MODEL is optional and can be left empty to use the provider default configured by the application.
Make sure Docker is running, then start the local MongoDB container:
bun db:upIf Docker reports that port 27018 is already allocated, another MongoDB container or service is already using that host port. Stop the conflicting service or change both MONGO_PORT and the port in MONGODB_URI.
bun run buildbun devKeep the terminal window open while using the application. Closing it stops the local frontend and backend.
Frontend:
http://localhost:5274
Backend API:
http://localhost:5002
Trending mode researches GitHub's own Trending page and turns the current ranking into a structured research report.
You can configure:
- Timeframe. Daily, weekly, or monthly
- Programming language. Optional
- Spoken language. Optional, using GitHub Trending's spoken-language filter
For each repository, GitHub Research can collect the repository name, description, programming language, total stars, forks, and the period-specific star count exposed by GitHub Trending. The report then identifies supported themes, notable repositories, the most represented language, and the top period-star mover.
Period-star counts come from GitHub Trending itself. GitHub Research does not attempt to reproduce or reverse-engineer GitHub's Trending ranking algorithm.
Emerging mode uses GitHub repository search to find recently created repositories that already meet the thresholds you choose.
Available filters include:
- Created within. Past day, past 7 days, or past 30 days
- Programming language. Optional
- Minimum stars
- Minimum forks. Optional
- Exclude forked repositories. Enabled by default
The resulting report focuses on current star counts, languages, repository topics, recurring clusters, and standout new projects.
In this tool, emerging means newly created repositories that match the selected filters. It is an operational research category, not a prediction that a repository will become broadly popular.
Keyword mode starts from a term or concept such as MCP, RAG, or browser agent and researches matching GitHub repositories.
The main search always keeps a current relevance landscape. You can optionally select a 7-day, 30-day, or 90-day recent repository window, which adds repositories created during that period instead of replacing the broader landscape with creation-only results.
Keyword research can use:
- Repository names and descriptions
- Programming languages
- GitHub topics exposed in repository search
- Current star counts
- An optional minimum-star threshold
- Recently created matching repositories when a recent window is selected
The LLM analyzes the normalized repository data and returns a research summary, recurring themes, notable repositories, and historical comparisons when previous matching runs exist.
Any research configuration can be saved as a recurring monitor.
Supported schedules include:
- Custom intervals in hours
- Daily
- Weekly
- Monthly
The local NestJS backend checks for due monitors every 15 seconds. Automatic runs occur only while the backend and MongoDB are running, so the included scheduler is intended for local or always-on deployments rather than offline execution.
The Monitors view shows saved configurations, next and previous run times, and the latest run status. Monitors can be run manually, paused or resumed, and deleted. Pausing a monitor stops future scheduled runs; it does not cancel research that is already in progress.
Each completed run is stored in History. When a matching previous snapshot exists, GitHub Research compares shared repository metrics and repository composition across runs. Comparison windows under six hours are treated as diagnostic snapshots rather than evidence of a durable trend.
GitHub Research is designed as a research layer rather than a replacement for every GitHub analytics tool.
| GitHub Research | GitHub Trending | OSS Insight | Star History | |
|---|---|---|---|---|
| Primary workflow | Research trending, emerging, or keyword-based repository activity | Browse repositories currently trending on GitHub | Query and analyze large-scale GitHub event data | Explore repository star growth and star-based trends |
| Keyword or concept research | Yes, with LLM-synthesized themes across matching repositories | No dedicated keyword-research workflow | Natural-language analytics queries over its GitHub dataset | Primarily repository and star-history oriented |
| LLM-synthesized repository themes | Yes | No | Analytics answers and visualizations rather than this report format | No |
| Emerging repository workflow | Yes, using creation windows and user-defined star/fork filters | No dedicated newly-created-repository mode | Rankings and data exploration | Star-growth and trending views |
| Recurring local monitors | Yes | No | Not the same local monitor workflow | No equivalent local research-monitor workflow |
| Stored reports and longitudinal comparison | Yes | No | Historical analytics are available | Star history is the core focus |
The comparison describes each product's main workflow. These tools solve overlapping but different problems.
| Variable | Description |
|---|---|
PORT |
Backend port. Defaults to 5002 |
PUBLIC_API_BASE_URL |
API URL used by the frontend |
PUBLIC_FRONTEND_URL |
Frontend URL used by the application |
MONGO_PORT |
Host port mapped to the MongoDB container |
MONGODB_URI |
MongoDB connection string |
DECODO_AUTH_TOKEN |
Decodo Web Scraping API Basic Auth token, without the Basic prefix |
DECODO_SCRAPER_ENDPOINT |
Decodo scraping endpoint. The included default is normally sufficient |
LLM_PROVIDER |
LLM provider: claude, openai, or gemini |
LLM_MODEL |
Optional model override |
ANTHROPIC_API_KEY |
Anthropic API key |
OPENAI_API_KEY |
OpenAI API key |
GEMINI_API_KEY |
Google Gemini API key |
The Settings page can change the active LLM provider and optional model override. API keys and the Decodo token remain environment variables and are never returned by the settings API.
| Mode | Setting | Options / behavior |
|---|---|---|
| Trending | Timeframe | Daily, weekly, monthly |
| Trending | Programming language | Optional |
| Trending | Spoken language | Optional GitHub spoken-language code |
| Emerging | Created within | 1 day, 7 days, 30 days |
| Emerging | Programming language | Optional |
| Emerging | Minimum stars | Numeric threshold |
| Emerging | Minimum forks | Optional numeric threshold |
| Emerging | Exclude forks | Enabled by default |
| Keyword | Keyword | Required |
| Keyword | Recent repository window | Any time, 7 days, 30 days, 90 days |
| Keyword | Programming language | Optional |
| Keyword | Minimum stars | Numeric threshold |
GitHub Research separates deterministic repository metrics from LLM interpretation.
- Deterministic facts stay authoritative. Repository counts, language counts, star counts, period-star counts, and source URLs come from parsed GitHub data rather than from the LLM.
- Themes require supporting evidence. The analysis prompt normally requires at least two current repositories before promoting a pattern into a theme. A single unusual project can still appear under notable repositories.
- Unknown names are not silently defined. If multiple repositories mention an unfamiliar product, acronym, model, or concept, the report can identify the observable cluster but should not infer what the term means unless the repository metadata establishes it.
- Emerging mode does not invent star velocity. GitHub repository-search results provide current star counts, not period-specific star gains, so Emerging reports do not claim a top period mover.
- Keyword windows are supplemental. A recent window adds recently created repositories to the current keyword landscape instead of treating every result as newly created.
- Historical claims use matching evidence. Repository deltas are compared only when the same repository and metric are available in earlier snapshots. Windows under six hours are explicitly treated as diagnostic.
| Method | Path | Description |
|---|---|---|
POST |
/research |
Run a GitHub research request |
POST |
/research/stream |
Run research with NDJSON progress events |
GET |
/queries |
List saved research history |
GET |
/queries/:id |
Retrieve one stored report |
DELETE |
/queries/:id |
Delete a history entry |
GET |
/monitors |
List recurring monitors |
POST |
/monitors |
Create a recurring monitor |
PATCH |
/monitors/:id |
Update, pause, or resume a monitor |
POST |
/monitors/:id/run |
Run a saved monitor immediately |
DELETE |
/monitors/:id |
Delete a monitor |
GET |
/settings |
Return provider/model selection and API-key presence status |
PATCH |
/settings |
Update provider or model preference |
Trending mode uses the repositories currently returned by GitHub Trending for the selected daily, weekly, or monthly window and optional language filters. GitHub Research analyzes that retrieved set and the period-star values GitHub exposes. It does not independently define or reconstruct GitHub's Trending ranking algorithm.
GitHub Trending provides a ranked view of repositories attracting attention for a selected period. GitHub Research can analyze that set, but it also adds Emerging and Keyword modes, LLM-synthesized themes, notable-repository explanations, stored reports, recurring monitors, historical comparisons, and Markdown or JSON export.
Keyword mode runs a current GitHub repository search for the supplied term and analyzes repository metadata such as descriptions, topics, languages, and stars. If you select a 7-day, 30-day, or 90-day recent window, the tool also retrieves repositories created during that period and adds them to the broader current landscape.
The current research flow uses Decodo Web Scraping API, not the GitHub REST API, to retrieve public GitHub pages. The backend then parses the returned HTML into normalized repository snapshots before passing structured data, deterministic facts, and relevant historical snapshots to the selected LLM provider.
GitHub documents a primary REST API limit of 60 requests per hour for unauthenticated requests and 5,000 requests per hour for most authenticated users, with separate or stricter limits for some resources such as search. GitHub Research does not use the REST API as its research data source, so those REST API request quotas are not the collection mechanism used by this project. See GitHub's rate-limit documentation.
Yes. GitHub Research retrieves its GitHub source pages through Decodo Web Scraping API, so you need a Decodo account and a valid Web Scraping API authentication token. The token is supplied through DECODO_AUTH_TOKEN in your local .env file and is not exposed by the application's Settings API.
No. GitHub Research is an independent open-source project maintained by Decodo. It researches publicly available GitHub repository data and links reports back to GitHub source pages, but it is not affiliated with, sponsored by, or endorsed by GitHub, Inc.
Yes. GitHub Research is designed to run locally or on an always-on host with Bun, MongoDB, and the required API credentials. Recurring monitors only execute while the backend and database are running, so scheduled monitoring requires the application to remain online between runs.
| Layer | Technology |
|---|---|
| Frontend | React 19, TanStack Router, TanStack Query, Tailwind CSS v4, Radix UI, Rsbuild |
| Backend | NestJS 11, MongoDB, Mongoose |
| GitHub retrieval | Decodo Web Scraping API |
| Parsing | Cheerio and deterministic GitHub parsers |
| LLMs | Anthropic Claude, OpenAI GPT, Google Gemini |
| Runtime | Bun |
| Local services | Docker Compose, MongoDB |
.github/
images/ # README screenshots and repository assets
apps/
backend/src/features/
decodo/ # Decodo Web Scraping API integration
github/ # GitHub URL builders and HTML parsers
research/ # Research execution, analysis, and reporting
monitors/ # Saved schedules and recurring execution
queries/ # History and stored snapshots
llm/ # Anthropic, OpenAI, and Gemini abstraction layer
settings/ # Provider/model configuration
frontend/src/features/
tracker/ # Research form, progress flow, and report UI
monitors/ # Recurring monitor controls
queries/ # History and stored reports
settings/ # Runtime provider/model settings
shared/src/ # Shared TypeScript types
docs/
ARCHITECTURE.md
| Command | Description |
|---|---|
bun dev |
Build shared types and start frontend/backend development servers |
bun run build |
Build all application packages |
bun start |
Start the production backend and frontend preview server |
bun lint |
Run linting across frontend and backend |
bun format |
Format frontend and backend source files |
bun db:up |
Start MongoDB through Docker Compose |
bun db:down |
Stop the local database container |
- Decodo Forum Scraper
- Decodo Stack Overflow Trends Monitor
- Decodo Web Scraping API
- Decodo SDK for TypeScript
- Decodo MCP Server
- Decodo OpenClaw Skill
For Web Scraping API targets, parameters, and request behavior, see the Decodo Web Scraping API documentation.
MIT – see LICENSE.