Sim

What Sim does

Sim is a workspace for building AI agents and the workflows around them. In a visual editor you connect blocks on a canvas: an agent block that calls a language model, blocks for apps such as Slack, Notion, HubSpot or Salesforce, conditions, loops and code. A workflow can start on a schedule, from a webhook, through an API call or from an app trigger, so it covers classic “when this happens in one app, do that in another” automation as well as agent tasks. Workflows can also be built by describing them in a chat or written in code.

The README lists more than 270 integrations with over 5,400 actions and 520 triggers. Besides workflows, the workspace holds knowledge bases that agents can search, structured tables and files, and every run is logged for monitoring. Several people can edit the same workflow at once, and a macOS desktop app can connect to the cloud or to a self-hosted server.

AI models and data

Agent blocks work with the major hosted model providers, using your own API keys, and with local models served by Ollama or vLLM. Running local models keeps prompts and documents on your own infrastructure, which is the main reason to self-host Sim rather than use the cloud. Knowledge bases store embeddings in PostgreSQL with the pgvector extension.

Some features depend on outside services even when you self-host. The built-in Chat assistant is a Sim-managed service that needs a Chat API key, and you can hide it if you do not want it. Code steps in plain JavaScript run locally in isolated V8 contexts. Python, shell and JavaScript that imports packages need a remote sandbox from E2B or Daytona.

Self-hosting Sim

An interactive setup wizard (npx sim-setup) creates a deployment folder, generates secrets and starts the published images with Docker Compose. It needs Node.js 20 or later and Docker. Manual Compose and a Kubernetes Helm chart are also documented. A deployment runs the app, a real-time server, PostgreSQL with pgvector, Redis and a scheduler. The documentation’s sizing suggestion for evaluation is 2 CPU cores, 12 GiB of RAM and 20 GiB of SSD storage, rising to 4 cores and 16 GiB for a team, which is heavy compared with most tools in this category. The production Compose file does not keep uploaded files on a volume, so configure object storage, and back up the encryption keys with the database.

License and editions

The root license is Apache-2.0, but the apps/sim/ee/ directory is under the proprietary Sim Enterprise License. It allows development, testing and internal non-production use, and requires a Sim Enterprise subscription for production use of features such as SSO, access control, audit logs, SCIM, whitelabeling and data retention policies. The self-hosting documentation describes switching these features on through configuration, so check the license before you rely on them in production.

Hosted version

Sim Cloud at sim.ai is run by the developers, and its MCP server lets assistants such as Claude or Codex run and manage workflows.

Limitations

Sim is AI-first, and for simple app-to-app automations it brings more machinery than you need. The self-hosted footprint is large, some features rely on Sim-managed or third-party services, and the enterprise directory makes the license less simple than the Apache-2.0 badge suggests.

Who it suits

Sim suits teams building AI agents that need to act in business apps, read internal documents and run on triggers, especially if they want the option of local models. For plain integrations between SaaS apps, Activepieces is lighter to run. The workflow automation guide compares the options.