Sim

What Sim is

Sim is an open-source AI agent workspace where teams build, deploy and manage AI agents for their organization. The homepage describes connecting 1,000+ integrations and every major LLM, then governing access, spend, data and deployment from one place, and building "visually, conversationally, or with code". The core is published under the OSI-approved Apache License 2.0 and supports self-hosting in your own cloud. That mix puts Sim in front of two readers at once: an operations or support team that wants a Slack bot or a lead-handling agent without writing an application, and an IT or platform team that wants agents run inside a governed workspace with permission groups, spend limits and audit records. It also fits the workflow automation and open-source low-code shelves, since every agent is a workflow and the core can be run on your own infrastructure.

What you build and how

The center of Sim is the workflow. The company describes workflows as repeatable agent systems that you design on a visual canvas, through a chat interface, or in code. On the canvas, Sim connects blocks, models and integrations into agent logic. The homepage examples show the pattern: a Start block, an Agent block prompted with a named model (GPT-6 Astra in one example, Claude Sonnet in another), a Slack block that posts a lead summary to a sales channel, and a Function block that runs code to format a support reply. Each block carries its own error path. When the visual blocks are not enough, code can be dropped into the flow, and the company's comparison guide says code and API capabilities let technical teams extend a workflow where a no-code block falls short.

There is also a command line. The CLI lets you use Sim from Claude Code or a terminal to build, run and manage agents; the homepage shows a sim workflows run command that feeds a ticket file into a workflow and follows the run. The footer's product list adds MCP, API and CLI access alongside Chat and Self Hosting, so the same agents can be driven from outside the canvas.

The data layer has three parts. The Knowledge Base gives every agent trusted memory from synced sources, and the homepage shows connectors for Notion, Google Drive, Confluence, Slack, GitHub and SharePoint with auto-sync. Tables store structured data that agents read and update between runs; the example is a leads table with name, company and a qualified flag. Files keep team uploads and agent outputs in one shared store, with folders such as brand assets and reports.

Logs trace every agent run block by block, including actions, cost and failures. The homepage mock-up shows a run overview with success rate, median run time, completed runs and cost over the last 24 hours, and a list of recent runs started by API, webhook or schedule. Those three trigger types are the ways a finished workflow gets called.

The governance layer is where Sim aims at IT teams. The homepage lists SSO and SCIM for identity-provider sign-in and automatic member provisioning, permission groups that control which models, integrations and tools each group can use, spend controls with organization and per-member usage limits and cost tracked by workspace, audit records of security-relevant changes, per-workspace data retention settings, and self-hosting with Docker or Kubernetes in your own cloud.

Named integrations in the site footer include Slack, GitHub, Gmail, Notion, Salesforce, Jira, Linear, Supabase and Stripe. The model list names OpenAI, Anthropic, Google, DeepSeek, xAI and Cerebras.

The Slack guide in Sim's library walks through a full chatbot build, which is the clearest picture of the product in practice. A Slack event or mention triggers a Sim workflow, the workflow checks identity and channel, retrieves approved knowledge, asks a model to compose an answer, optionally calls read-only tools or routes a sensitive action to human approval, and replies in Slack. The guide covers Slack app setup, permissions, events, knowledge sources, actions, testing, deployment, security and maintenance. It also notes a limit: slash commands require a custom Slack bot configured with its signing secret, because the native Sim Slack app does not subscribe to slash commands. And it is candid that "Slack app configuration and security decisions are still required even when the workflow itself is visual."

Licensing, plans and security

The licensing is spelled out in the library. Sim's core is open source under the Apache License 2.0 and supports self-hosting. Features in the enterprise folder, which the guides list as SSO, SCIM, access control, audit logs and white-labeling, use a separate Sim Enterprise License. That license is free for development, testing and internal non-production use, requires an Enterprise subscription for production use, and does not permit modification or redistribution.

For the hosted product, the company's comparison guide says Sim offers a managed cloud option with credit-based usage. The pages read here print no plan prices, and the library itself says hosted-product terms and pricing should be verified separately, because a software license does not define cloud-service pricing. The FAQ adds that the core can be self-hosted without software license fees, while infrastructure, model, database and external service costs remain the operator's.

The homepage shows three trust badges, SOC 2 Type II, ISO 27001 and GDPR, each with details in the Sim Trust Center. Customer stories feature Rivian and eXp Realty. The homepage counters claim $20,000,000 saved by teams using Sim to automate their work, 1,270,000 hours of work completed in Sim, and 100,000+ builders using Sim to create and manage AI agents. The GitHub link in the header shows 29.8k. Resources include docs, a blog, a changelog, a status page, a security page and the library of guides.

Who it suits and who it doesn't

Sim fits technical teams that want both a visual builder and an escape hatch into code, and that care about where agents run. An operations team building an internal Slack assistant grounded in Confluence or Notion, a support team routing tickets through a model with human approval on sensitive actions, or a sales team enriching inbound leads into a shared table are all close to the homepage's own examples. An IT team that needs SSO, SCIM, permission groups, spend limits and data retention gets them in one workspace, with the option to run the core on its own Docker or Kubernetes setup.

The caveats come from Sim's own pages. Self-hosting moves responsibility for infrastructure, upgrades, security patches, monitoring, backups and model or API costs to the operator, and the comparison guide says plainly that it is not automatically cheaper than a managed service. The governance features that enterprises want sit under the Enterprise License, so production use of them is a paid subscription even on your own servers. A solo founder who only wants a simple website chat widget will find more machinery here than the job needs, and the Slack guide's long list of scopes, events and approval gates shows that a safe bot still takes deliberate setup work.