Sixb
docs.sixb.ai- Category
- Developer Tools
- Rank
- No. 1231Tools index
Previous survey · No. 1242 ·
- Type
- TOOL
- Use case
- Data, Retrieval & Knowledge
- GitHub
- 136 stars
- Latest release
- v0.2.2
- Date
About
Sixb is a TypeScript framework for defining a shared "ontology" of objects, properties, links, and value types that connects data sources like Pipedrive, PandaDoc, and Google Drive into a single typed model. Applications, AI agents, and human-in-the-loop workflows can then query and act on that shared model, with actions able to write results back to the original connected systems.
At a glance
Sixb is a TypeScript framework for building apps and agents around a shared model of your business data. It is worth prototyping if a UI, an agent and an approval workflow all need to work with the same records and actions.
Good fit
Teams connecting several sources to an internal operations app that reads data and performs reviewed actions.
Consider something else when
Skip it for a small app that only needs to read a single table or service.
Start here
Scaffold a local project, then model two synthetic customers and one quote. Query the quote in the app and build a mock approval step. Check that approving it updates the sample state before wiring up external actions.
Cost and requirements
The documented quickstart requires Bun 1.4.4 or later and needs no database setup or API key. Real deployments can add infrastructure, connector credentials and model costs.
Limits to know
Connected systems need their own credentials and data mappings. Actions can write back to those systems. Start with sample data and confirm that roles and approval rules cover the writes you intend to expose.
Sources checked 2026-11-05. Source-based assessment.
Read the source review
What it does
You describe the things your business deals with, such as quotes, customers and invoices, once in TypeScript. Sixb pulls rows from the software you already run into versioned tables, cleans them with SQL, and maps them onto those types. One generated project then serves a React app, an HTTP API with its own reference page, a terminal client and a chat agent, all reading the same records. Changes go through typed actions that check permissions and can push the update back to the originating service.
Why it's ranked here
The scope is unusually wide for an open source project, and the code backs it. The command line can run the API, background workers, a scheduler, an orchestrator and a rules engine as separate processes, and one workspace holds pluggable families of connectors, model providers, sandboxes, queues and storage. The pitch covers business tools such as Pipedrive, PandaDoc and QuickBooks. It is MIT licensed. The catch is maturity: its own readme says early releases can break things and change the database.
What's good
Setup is light: a single create command scaffolds a project that runs locally with no database to provision and no API keys. Project structure is by convention, one folder per stage from connection to app, so the layout tells you where things go. Every app ships with an admin console showing each sync, pipeline, projection and action run alongside its logs, plus settings pages for members, tokens and AI usage. The agent runs inside a sandbox and, in chat, carries only the permissions of the person asking.
Tradeoffs
It is tied to Bun: the terminal client refuses to start on older versions and the manifests target a recent Bun release. The readme warns that the early release line can include breaking changes and database migrations, so upgrades need reading the changelog first. The surface area is large, with a long list of authorization, event and sharing concepts in the core library, which is a lot to learn before the payoff. It is a whole application platform, not a library you drop into an existing app.
How to use it well
Fit it to a small team that runs its operations across several business tools and wants one typed model with an internal app and an agent on top. Start from the scaffold and the reference app before modelling your own domain, and keep a project on a pinned release while the API is still moving. If you only need a data pipeline or only a chat agent over a database, this is more framework than the job needs.
Technical notes+
package.json declares a Bun workspace (packageManager bun@1.4.4) spanning packages, connectors, models, sandboxes, storage, broker, queues, auth, env, loggers and deploy, with tests run through bun test and linting through Biome. packages/core/src/index.ts exports the authoring API (defineObjectType, defineAction, defineDataset, defineRule, defineRole, defineMarking) alongside event, predicate and authorization types. packages/server/src/server.ts builds the API on Elysia with an OpenAPI plugin, CORS limited by a browser-origin policy, CSRF and shared-access headers, run-token checks resolved before body parsing, and a 64 KB cap on WebSocket messages. packages/cli/src/index.tsx enforces a minimum Bun version and dispatches dev, worker, api, agent, db:migrate and lake:cleanup subcommands. sandboxes/local/src/index.ts exposes a local sandbox factory with isolation detection.
Observed
- License
- MIT (LICENSE)
- Language
- TypeScript
- Runtime
- Bun; CLI exits on versions below its minimum (packages/cli/src/index.tsx)
- Install surface
- bun create sixb scaffolder; Bun workspace monorepo (package.json)
- Interfaces
- TypeScript library, HTTP API with OpenAPI docs, CLI, React app shell, built-in agent
- Server framework
- Elysia with CORS and OpenAPI plugins (packages/server/src/server.ts)
- Stability notice
- Readme states early releases can include breaking changes and database migrations
- Test tooling
- bun test and Biome lint scripts declared in package.json
Read from README.md, package.json, packages/core/src/index.ts, packages/agent/src/index.ts, packages/orchestrator/src/index.ts, packages/server/src/server.ts, packages/cli/src/index.tsx, sandboxes/local/src/index.ts, connectors/github/src/index.ts, docs/README.md, docs/apps/overview.md, LICENSE.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.