Sim AI: An Open-Source Workspace for Building and Deploying AI Agents

What Sim is and the problems it targets

Sim is a workspace for building, deploying, and managing AI agents and workflows. The README describes it as "the central intelligence layer for your AI workforce". In practical terms, it addresses a common infrastructure problem: teams building AI-powered products need to connect language models to external services (Slack, Notion, HubSpot, Salesforce, databases), manage conversation history, handle file and knowledge-base ingestion, and schedule or monitor agent runs. Sim packages all of that into a single deployment rather than assembling it from separate services. Agents can be built visually through a drag-and-drop interface, conversationally through a chat interface, or programmatically with code. The README lists files, tables, and knowledge bases as first-class objects that live in the same workspace alongside agents and workflows.

Three deployment modes: cloud, desktop, and self-hosted

Sim is available in three modes. The cloud-hosted version at sim.ai requires no installation. The macOS desktop app (Apple Silicon and Intel, macOS 12 or later) runs locally and can point to a self-hosted deployment. The self-hosted mode uses Docker Compose and is the focus of the open-source repository. The README states that self-hosting requirements are Node.js 20 or later and Docker. A warning in the Cursor prompt link notes that the stack needs 12 GB of RAM or more. Each mode connects to the same workspace model: agents, tables, files, and knowledge all share the same data layer regardless of where Sim is running.

Self-hosting with npx sim-setup

The `npx sim-setup` wizard is the primary self-hosting path. It creates a deployment directory, provisions the database, generates secrets, writes `.env`, connects a Chat API key, and starts the published Sim Docker images. When the wizard finishes, the app is available at http://localhost:3000. The wizard does not clone the repository; it uses the published images directly. To add optional capabilities after the initial setup:

npx sim-setup config
npx sim-setup add email
npx sim-setup add storage
npx sim-setup add knowledge
npx sim-setup add integration slack

Ongoing management uses the same tool:

npx sim-setup start | stop | restart
npx sim-setup update
npx sim-setup status
npx sim-setup logs
npx sim-setup doctor

The `--dir <path>` flag creates or manages a deployment at a path other than the default `./sim`. Local model support via Ollama and vLLM is available as a Docker Compose variant documented at docs.sim.ai/self-hosting/docker.

Chat API keys and the connection to sim.ai

One aspect of the self-hosted path that requires an external connection is the Chat API key. The README states that Chat is a Sim-managed service; `npx sim-setup` opens a browser to sign in and store the key automatically. Keys can be viewed, created, or revoked at sim.ai/selfhost/settings/chat-keys. This means the self-hosted deployment is not fully air-gapped: it needs an account at sim.ai for the Chat feature. For teams that cannot accept any outbound dependency on a third-party service, this is a real constraint. The full environment variable reference is at docs.sim.ai/self-hosting/environment-variables, and defaults are documented in `apps/sim/.env.example` within the repository. The Cursor prompt link embedded in the README shows the required secrets: BETTER_AUTH_SECRET, ENCRYPTION_KEY, INTERNAL_API_SECRET, CRON_SECRET, and POSTGRES_PASSWORD, each generated with `openssl rand -hex 32` or `openssl rand -hex 24`.

Tech stack and source structure

The project is a Bun monorepo using Turborepo. The package.json lists the package manager as `[email protected]` and workspaces as `apps/*` and `packages/*`. The tech stack listed in the README details section is: Next.js (App Router) for the frontend, Bun as the runtime, PostgreSQL with Drizzle ORM for the database, Better Auth for authentication, Zod for schema validation, and Shadcn with Tailwind for the UI. The docker-compose.prod.yml is the file used by the `npx sim-setup` installer for production deployments. An Ollama variant is available in docker-compose.ollama.yml. A Helm chart in the `helm/` directory supports Kubernetes deployments, which the README notes is available when running `bun run sim-setup` from inside a cloned repository. Kubernetes mode is not accessible via the npx path.

Limitations and when Sim is not the right choice

Sim requires Docker and Node.js 20 for self-hosting, plus at least 12 GB of RAM. The Chat feature requires an active sim.ai account and outbound connectivity to sim.ai, which rules it out for fully air-gapped deployments. The version in package.json is 0.0.0 with the package marked private, suggesting the npm packages are not published individually; the deployment relies on Docker images. For teams that need only a thin LLM integration layer rather than a full agent platform, the overhead of running a PostgreSQL database, a Bun server, and a Next.js front end may be excessive. LangChain and LlamaIndex are lighter-weight Python alternatives that focus on building LLM-backed workflows in code rather than through a visual workspace, with no self-hosted web UI and therefore lower infrastructure overhead. Sim's advantage over those is the visual builder, the integrated tables and knowledge store, and the first-class multi-agent orchestration UI.

Editorial conclusion

Sim suits teams that want to build AI agent workflows without writing a bespoke orchestration layer from scratch. The cloud version at sim.ai is free to try; the self-hosted path via npx sim-setup requires Node.js 20, Docker, and approximately 12 GB of RAM. It is a poor fit for environments where all data must stay fully offline and the team cannot connect a Chat API key to sim.ai for the deployment, or for single-developer projects that need only basic LLM prompting rather than a full agent-and-workflow platform. The project released v0.9.4 on 2026-09-28 and the last push was on 2026-09-26, indicating fast iteration.

Frequently asked questions

Official sources

1. Official documentation
2. Official README
3. Project repository
4. Release notes