Get the weekly radar
One short email with the freshest products, every week.
Sim is an open-source AI workspace that builds agents from a plain-English description.
Grounded in available product and source data
Wiring an AI agent together by hand, tool by tool, is exactly what Sim is built to shortcut — it's an open-source AI workspace for building agents where describing what should happen in plain English is often enough to get a working setup, per the site's own framing: 'Describe it. Sim builds it.'
The workspace splits into five pieces that map onto how an agent actually gets built and kept honest: Workflows for designing agent logic visually, a Knowledge Base giving agents semantic memory of a team's own data, Tables for structured data the agent can read and write, Files as one shared store for both people and agents, and Logs tracing every decision an agent made, so a run can be audited rather than trusted blindly.
A support-triage example on the site shows the pieces working together: an unread customer email comes in, gets run through a Claude model with help-center knowledge attached, escalates to an on-call engineering channel if it's urgent, drafts a friendly reply, and logs the ticket — assembled from blocks rather than hand-written integration code for each step.
Sim connects to a broad, explicitly named stack — Slack, GitHub, Gmail, Notion, Salesforce, Jira, Linear, Supabase, and Stripe among 1,000+ integrations — and works with OpenAI, Anthropic, Google, DeepSeek, xAI, Cerebras, Groq, and Sakana AI rather than locking into a single model provider. Self-hosting is offered alongside the hosted version, and the site states it's SOC2 compliant and trusted by over 100,000 builders, positioning itself against workflow tools like n8n, Zapier, Make, and Retool rather than against other agent frameworks specifically.
Not necessarily — an agent can be built by describing what it should automate in plain English, with Sim assembling the blocks, models, and integrations into a workflow that's still reviewable, customizable, and testable in the visual builder afterward.
They act, not just read — the support-triage example on the site shows an agent escalating to a Slack channel, replying to a customer, and writing a row to a table in one automated pass, not simply summarizing an inbox.
No — the site lists OpenAI, Anthropic, Google, DeepSeek, xAI, Cerebras, Groq, and Sakana AI as supported models, so a workflow isn't tied to a single vendor's model by default.
No — self-hosting is offered as an option alongside the hosted version, which matters for teams that need to keep agent infrastructure inside their own environment.
The site positions itself specifically for AI agents rather than general trigger-action automation — Knowledge Base, semantic memory, and per-block decision logging are built around agent behavior, which is also why the site names those tools directly as comparison points rather than avoiding the comparison.
These 16 comments were collected from the discovery source. ProductVisit does not host replies in Release A.
👋 Hey Product Hunt! I'm Emir, co-founder and CEO of Sim. Today Sim is open to everyone, and I couldn't be more excited to share it ❤️ Sim started with a mess of our own making. Waleed (my best friend and co-founder) and I were prompting Claude to build sophisticated automations in n8n, storing data in Supabase, and standing up infra for our APIs and MCPs, and we realized the stack we'd assembled just to build agents that automated our own work was a complete mess. Not to mention our token spend... So we set out to build the one platform we wished we had. The problem: building AI agents today means stitching together frameworks, one-off scripts, and brittle automations that break the moment anything changes. We wanted one place to build an agent, give it access to our data across 1,000+ integrations, build a brain for memory, deploy it, and actually manage it over time. So we built one. Sim is the open-source AI workspace for agents. Here's what makes it different: 🗣️ Build by chatting — describe what you want and Sim builds the agent and workflow for you. Or design it visually on a canvas. Or drop into code. Whatever fits the job. 🔌 1,000+ integrations + every major LLM — Slack, Notion, HubSpot, Salesforce, Gmail, and more, connected out of the box. Your agent says "message me on Slack when a deal closes" and it just works. 🧠 One workspace, shared context — Workflows, Tables, Knowledge Bases, and Files all live together, so your agents share memory and data instead of living in disconnected tools. 💸 Cost-efficient by design — Sim swaps token-hungry tool calls for deterministic steps and real code wherever it counts, so you're not burning tokens (and money) on work that never needed an LLM in the first place. 🚀 Built for real work — Slack bots, compliance agents, data pipelines, research assistants. Not demos, actual production agents. 🔓 Open source (Apache 2.0), SOC2, and already trusted by 100,000+ builders. Who it's for: teams who want to put AI agents to work (IT, ops, and technical teams who need governance and control), and individual builders who care about speed and open source. We're shipping fast and want to build this with you. Tell us what's missing, what's broken, and what would make Sim 10x more useful for you. Try it -> sim.ai I'll be here all day and will read and reply to every single comment 🙌 Show more
10 source upvotesThe "chat to build" vs canvas vs code approach is interesting, most tools force you to pick one. Which one do most of your users actually end up sticking with?
2 source upvotesCongrats on the launch! 100k builders on an open source tool is genuinely impressive. BTW, what's the split between solo devs and actual teams using this in production?
2 source upvoteswho do you think gets the most value out of Sim today, developers, technical teams, or can non-technical users get productive quickly as well?
2 source upvotes@emirkarabeg The mess of stitching frameworks and watching token spend go wild is incredibly relatable. Sim solving this by bringing workflows, tables, and knowledge bases into one shared context is huge for real production work. Can we seamlessly toggle between the no-code canvas and dropping into raw code for specific nodes, or do we have to pick one style for the workflow?
2 source upvotesLove the emphasis on replacing unnecessary LLM calls with deterministic code instead of assuming everything needs an agent. Just Curious have you noticed users converging toward a small set of reusable agent patterns over time, or is every team's workflow still highly bespoke? Congrats on the launch! 🚀
2 source upvotes@abod_rehman Thanks so much! Great question. Solo agent builders account for 70%+ of the users on the platform, while teams actually account for 90%+ of agent and workflow runs.
1 source upvotes@boyuan_deng1 Originally, Sim users were only building on the canvas. Now, more than 90% of the platform's usage comes from the chat. Sims love prompting!
1 source upvotes@aymi_malik Great question. Technical teams building agents and solo devs looking for automations get the most value out of Sim today. Non-technical users still get loads of value from the Sim chat.
1 source upvotes@emirkarabeg Being able to interchange it node-by-node gives the perfect balance between speed and full control. Absolute game-changer for dev workflows. Thanks for clarifying.
1 source upvotesHow does Sim actually compare to n8n or Langflow for someone who already has a few agent flows running? Curious what the real differentiator is beyond the open-source angle.
1 source upvotes@gkeaksoyakhabw Sim's biggest differentiator is the ability to build across your entire workspace with chat. You're able to monitor and debug production workflows with Sim and have it alert you for any errors or anomalies in production agents. Sim has agent skills, file storage, native tables and vector storage, and some more capabilities that are built natively alongside workflows. See a full list of comparisons here: https://www.sim.ai/comparison
1 source upvotes@emirkarabeg Congratulations on the launch! "1,000 integrations" that's massive🙌🏽
1 source upvotes@kutlwano_melamu thank you!
1 source upvotes@tehreem_fatima5 you can decide which option you'd like to work with and interchange for each node in your workflow.
@tarqiya_forgah A pattern we're seeing often is when teams put a set of deterministic nodes first (fetching data, cleaning it, etc.) and then put the LLM-based nodes at the end to handle the last-mile and other nuances.