Graphify

Quick Verdict

Graphify is an open-source code knowledge graph maintained by Graphify Labs. The current official identity is precise: the product remains Graphify, the canonical website is graphify.com, the source repository is Graphify-Labs/graphify, and the PyPI package is graphifyy with two y’s. It is not another AI editor or a hosted vector database. It sits beside an existing coding assistant, converts code relationships into a local graph, and lets the assistant traverse that graph instead of starting every question with broad text search.

Its strongest design choice is visible provenance. Tree-sitter parses code relationships locally and labels them EXTRACTED; a user-selected model may connect non-code sources and labels those edges INFERRED; unresolved evidence is marked AMBIGUOUS. Graphify is worth testing if Claude Code, Cursor, or GitHub Copilot repeatedly has to rediscover a large repository. For a small project where built-in search already answers most questions, maintaining another generated artifact may not be worthwhile.

Best For

• Developers maintaining medium or large repositories who need call paths, dependency direction, and change impact rather than same-text matches.
• People who switch among coding assistants and want a reusable context layer outside any single editor or model vendor.
• Platform and architecture teams that inspect critical nodes, module communities, dependency cycles, or pull-request impact with file-and-line evidence.
• Privacy-conscious local workflows where the team can manage a Python CLI and distinguish deterministic local parsing from optional model inference.
• Not ideal for single-file projects, users seeking a complete graphical IDE, or teams unwilling to maintain installation versions, ignore rules, and graph updates.

Key Features

• Local AST graph construction: official documentation lists Tree-sitter parsing for 36 code languages, extracting files, functions, classes, methods, imports, and calls without a model request.
• Traceable graph queries: graphify query, graphify path, and graphify explain explore connections and return evidence tied to source files and lines.
• Incremental workflow: /graphify . creates a graph, while --update refreshes it; graph.json, graph.html, and GRAPH_REPORT.md are stored under graphify-out/.
• Optional non-code inference: docs, PDFs, SQL schemas, PostgreSQL, and Terraform can be connected by configured Claude, OpenAI, Gemini, DeepSeek, Kimi, Bedrock, Azure, or local Ollama backends.
• CLI and MCP access: the same graph can be queried directly or served over stdio or HTTP through ten MCP tools for nodes, neighbors, paths, communities, graph statistics, and pull-request impact.
• Assistant integrations: Graphify documents setup for 17 coding assistants, including Claude Code, Cursor, Codex, Gemini CLI, Aider, GitHub Copilot, and OpenCode.

Use Cases

When onboarding to an unfamiliar repository, a developer can build the graph and ask how authentication reaches the database, which modules call a service, or where a symbol sits in the dependency chain. During refactoring, path and neighborhood queries can reveal a likely blast radius before code changes begin. Pull-request tools can then map changed files onto existing relationships to support review and conflict triage.

Graphify also fits projects where code, architecture notes, database definitions, and infrastructure files must be considered together. The provenance labels matter here: an AST-derived call edge is not equivalent to a model-inferred link from a PDF. Teams should preserve the labels and verify important conclusions against source, tests, and database definitions rather than treating every graph path as proven fact.

Pricing

As of July 21, 2026, Graphify’s official pricing page described the core product as free, MIT-licensed, account-free, unlimited, and requiring no payment card. This directory therefore classifies it as free rather than freemium. The official install command is uv tool install graphifyy, with pipx and pip also supported. Similar PyPI names are unrelated, so package identity should be checked before installation.

Graphify Labs separately advertises an enterprise early-access track with merge-gate verification, graph-aware review, an engineering digest, and self-hosting inside a customer’s VPC. No public enterprise price was available at the cutoff. Organizations should request a quote and confirm support boundaries, deployment responsibility, security evidence, and contract terms rather than assuming that enterprise capabilities share the core tool’s free license.

Pros

• Deterministic code parsing stays on the local machine and does not require a model or hosted index.
• EXTRACTED, INFERRED, and AMBIGUOUS labels keep parsed facts, model judgments, and uncertainty separate.
• CLI, MCP, and 17 assistant integrations make one graph usable across development tools.
• Outputs are inspectable local files rather than a mandatory vendor-hosted private index.
• The repository uses the permissive MIT license and showed same-day releases and active fixes at the research cutoff.

Cons

• Requires a Python toolchain plus team decisions about versions, ignore rules, generated output, and refresh timing.
• Non-code sources may pass through the selected model backend; a cloud backend introduces that provider’s data handling terms.
• Static analysis cannot fully resolve reflection, dynamic dispatch, generated code, or runtime-only dependencies, so ambiguous edges need review.
• Small repositories may be served more directly by editor search or existing indexing without the extra build step and disk artifacts.
• Enterprise is still early access with no public pricing; the official security page said SOC 2 Type II was in progress, not completed.

Alternatives

These products are not mutually exclusive. Choose a coding assistant when the missing capability is task execution and code editing; choose Graphify when the missing layer is persistent, traversable repository structure that several assistants can query.

FAQ

What are Graphify’s canonical website and repository?

The canonical website is https://graphify.com/, and the source repository is https://github.com/Graphify-Labs/graphify. The former graphifylabs.ai domain now explicitly points to the new canonical site. The product name and directory slug remain Graphify, so this is an ownership and URL update, not a new product.

Is Graphify free?

Yes. The core CLI and knowledge-graph tool are free, account-free, and MIT-licensed with no public usage limit. Enterprise early-access features have no published price and require contacting Graphify Labs.

Does Graphify upload my source code?

Tree-sitter code parsing, graph files, and core queries run locally, and Graphify says the core has no telemetry. Optional non-code inference is handled by the backend you configure. Local Ollama can keep that path local; a cloud model requires review of that provider’s data policy.

Does Graphify require an API key?

Building the deterministic code AST graph requires neither a Graphify account nor a model API key. Cloud-backed inference for docs, PDFs, or schemas normally requires that provider’s credentials, while local Ollama does not require a cloud API key.

How is Graphify different from RAG or a vector database?

Graphify’s code layer stores AST entities and typed relationships, then follows calls, imports, and references. Conventional RAG usually retrieves chunks by semantic similarity. Graph traversal is better suited to dependency and blast-radius questions, while similarity search is useful for finding related prose; they can be combined.

Can Graphify replace Cursor, Claude Code, or GitHub Copilot?

Not completely. Graphify supplies structured repository context and query tools; it is not a full editor or general coding agent. Its typical role is to complement one or more of those assistants.

What are Graphify’s license and security-certification status?

The public repository’s root license is MIT, with copyright assigned to Safi Shamsi in 2026. At the page’s research cutoff, the official security page described SOC 2 Type II as in progress for the enterprise track, not as a completed certification. Enterprise users should still assess dependencies, model backends, deployment, and support terms.

Bottom Line

Retain the canonical slug graphify. The product identity did not change: the official website moved from graphifylabs.ai to graphify.com, and the repository now belongs to the Graphify Labs organization. On July 21, 2026, the project published v0.9.23 and continued merging fixes, which rules out archival on inactivity grounds.

Graphify is a credible option for developers who already have an AI coding assistant but lack persistent, queryable code relationships. Start with the free MIT version on one real repository, inspect its citations and update behavior, and map every optional model call to an acceptable data path. Keep it only if graph queries repeatedly reduce rediscovery and improve change-impact decisions.