hashintel/hash

What is HASH?

HASH is an open‑source, self‑building database that turns raw information into structured knowledge graphs (called world models). It continuously ingests data from public web sources and private integrations, validates and links it, and makes the resulting entities and schemas searchable and usable by both humans and autonomous AI agents.

Core ideas

• Graph‑backed knowledge – data is stored as nodes (entities) and edges (relationships) with explicit schemas (types).
• Bitemporality – every fact records when it happened and when it became known, enabling provenance‑aware queries.
• Autonomous agents – built‑in agents can crawl the internet, pull from APIs, or sync private sources, automatically populating and checking the graph.
• Self‑checking – the system validates schema conformance and consistency as new data arrives.
• High‑trust source of truth – the graph can be queried by downstream applications for decision‑making, automation, or optimisation.

Repository layout

Folder	Purpose
apps/	Main services (API, frontend, auth, workflow engine, etc.) that together run the HASH platform.
blocks/	Public Block Protocol blocks – reusable data‑model components that can be shared across projects.
infra/	Docker‑Compose files and deployment scripts for running the whole stack locally or self‑hosting.
libs/	Shared libraries – npm packages and Rust crates used by the services.
tests/	End‑to‑end and integration test suites.

Getting started

Quick‑start (recommended)

1. Create a free account at hash.ai and sign in. The hosted service spins up a ready‑made instance in under five minutes.

Run locally (for developers)

1. Install Git, Docker Compose, Docker Buildx, and mise (a version manager). Give Docker at least 8 GB RAM.
2. Clone the repo and install pinned toolchains:

   git clone https://github.com/hashintel/hash.git && cd hash
   mise trust && mise install
   yarn install
   

3. Create a .env.local with dummy API keys (only needed for AI‑powered features).
4. Bring up the supporting services:

   yarn compose up -d   # Postgres, Redis, Kratos, Hydra, Temporal, Vault, MinIO, Grafana, etc.
   

5. Start the platform:

   # Fastest for development – compile the Rust graph once, then run API + frontend with hot‑reload
   yarn start:graph   # in terminal 1
   yarn dev           # in terminal 2
   

   The UI becomes available at http://localhost:3000.

If you don’t need to compile the Rust graph, you can run it in Docker with the hgres profile.

Main components

• API (hash-api) – REST/GraphQL entry point, handles authentication (via Kratos), entity CRUD, and workflow orchestration (Temporal).
• Frontend (hash-web) – React‑based UI for browsing entities, editing types, and managing agents.
• Graph engine – Rust implementation that stores the knowledge graph in PostgreSQL and provides fast query capabilities.
• Temporal workflows – background workers that run autonomous agents, data ingestion pipelines, and consistency checks.
• Auth stack – Ory Kratos + Hydra for user management and OAuth/OIDC.
• Observability – Grafana dashboards, Prometheus metrics, and logs for debugging.

Extensibility & integrations

• AI providers – OpenAI, Anthropic (keys set via env vars) power agent reasoning and summarisation.
• File storage – Local filesystem by default; optional AWS S3 via FILE_UPLOAD_PROVIDER.
• Email – Dummy transporter for dev, or real SMTP / AWS SES for production.
• Block Protocol – reusable schema blocks can be imported from the public Þ Hub.

License

Most of the code is released under permissive open‑source licences (see LICENSE.md). Some sub‑components may have different terms, all listed in the license file.

Community & support

• Discussions & issues – GitHub Discussions and the issue tracker.
• Social – Follow @hashai on X (formerly Twitter).
• Hiring – The company posts open roles on its website; contributions are a way to get noticed.

---

In short: HASH is a full‑stack, open‑source platform for building and maintaining a live, provenance‑rich knowledge graph that can be queried by humans and AI agents alike. The repo provides everything needed to run the service locally, self‑host it, or use the hosted SaaS version.