Neutral · GitHub-backed · agent-queryable
GraphCanon
The stack decision graph for AI developers
Not "top 10 by stars." Constraint-first comparisons, typed OSS relationships, and honest when-not-to-use guidance - for humans and the agents that pick tools at runtime.
Browse by category
AI Agentshermes-agent · ECC · AutoGPT727 →Computer Visiontransformers · ai-engineering-from-scratch · LocalAI158 →Data & Retrievalgenerative-ai-for-beginners · graphify · awesome-llm-apps396 →Developer ToolsECC · prompts.chat · OpenHands639 →Evaluation & Observabilitygenerative-ai-for-beginners · headroom · CL4R1T4S440 →Inference & Servingtransformers · open-webui · ollama407 →LLM FrameworksAutoGPT · transformers · tensorflow503 →Model Trainingtransformers · tensorflow · LLMs-from-scratch563 →Speech & Audiotransformers · LocalAI · whisper.cpp205 →Vector Databasesmilvus · WeKnora · qdrant151 →Popular decisions
Head-to-head pages developers search for.
langchain vs llama_indexlanggraph vs crewAIqdrant vs chromaqdrant vs weaviateqdrant vs pgvectorqdrant vs milvus
All comparisons →Trending tools
Curated stacks
How the categories fit together to build something real.
For agents
Every page has a .md twin and JSON. Traverse the graph or resolve tools by constraints over MCP.
Common questions
- What is GraphCanon?
- GraphCanon is a knowledge graph of open-source AI development tools - frameworks, vector databases, agent runtimes, and related libraries - with typed edges, compare pages, and machine-readable JSON. We model the catalog as a Schema.org Dataset (https://schema.org/Dataset) so crawlers and agents can discover the full graph programmatically.
- How do tools get listed?
- Anyone can suggest a GitHub repository via Submit. We fetch public metadata on ingest, generate a neutral summary, assign categories and tags, and propose graph edges against existing tools. Signed-in submissions publish immediately; anonymous suggestions queue for review.
- How do compare and alternatives pages work?
- Compare pages pit two published tools side by side with live GitHub stats, shared categories, and typed graph relationships. Alternatives pages rank similar tools from the graph (alternative and successor edges). Neither is a paid placement or a single winner score.
- What do trust signals mean?
- Trust signals show sourced, dated heuristics from public GitHub metadata and optional dependency scans - maintenance label recency, provenance, and scan findings. They are decision inputs, not security certifications. Read trust methodology for methodology and explicit limits.
- How can AI agents query GraphCanon?
- Every HTML page has a markdown twin. Use the llms.txt convention for background, fetch /llms.txt for the live link index, POST MCP endpoint for Model Context Protocol tools, or the open REST API at GET /api/graphcanon - no API key required for read access.
- Is GraphCanon neutral?
- We do not accept payment for rankings. Stats come from GitHub and our ingest pipeline; summaries and edges are generated under documented methodology versions. You should verify fit for your constraints before adopting any tool.