Coworker vs Sim (2026)

About thirty five connectors are named on the platform and agents pages, out of a stated 50+ OAuth connectors listed in a directory at /connectors.

On the platform page they cover CRM (Salesforce, HubSpot), communications and meetings (Slack, Gmail, Zoom, Google Meet, Google Calendar, Microsoft 365), docs and project work (Google Drive, Google Sheets, Notion, Confluence, Jira, Linear, Asana, Figma), engineering (GitHub, Datadog, PagerDuty, VS Code), data (Snowflake, BigQuery, Amplitude), payments (Stripe), the browser (Chrome) and the systems of record for IT service management, HR and finance (ServiceNow, Workday, ADP, SAP, NetSuite).

The agents page adds Gong for sales, Zendesk and Intercom for support, and Microsoft Teams. Every connector reads and writes, so agents pull data and push updates across CRM, comms, support, docs, code and data, updating CRM records, creating tickets, sending follow-ups and drafting reports, and the published run drafts into Gmail, updates Salesforce and posts to Slack.

Agents inherit the permissions already in place, with no new logins and no new SSO maps, so there is no parallel access model to drift. Chrome and VS Code appear in the OAuth connector list as ingestion paths feeding the graph, not as interface control.

Sourcecoworker.ai/platform integrations section and /agents connector listread 2026-09-12

More than 1,000 native integrations span communication (Gmail, Slack, Teams, Telegram, WhatsApp), productivity (Notion, Google Workspace, Airtable), development (GitHub, Jira, Linear), CRM and revenue (Salesforce, HubSpot, Attio, Gong), data and search (PostgreSQL, BigQuery, Pinecone, Perplexity, Exa) and security and IT (CrowdStrike, Okta, AWS IAM), exposed to agents as tools inside workflows. Anything not built in connects through MCP.

Sourcedocs.sim.ai/introductionread 2026-09-29

The Agent Builder is the authoring surface, where customers create custom agents in minutes by defining what they do, which tools they use and when they run, with no code required, and set triggers, approval gates and notification preferences through a visual interface. Agent sequences span several steps and systems, and they branch.

In the account agent's published run, it drafted outreach for six accounts through Salesforce and Gmail, escalated one strategic account to an AE through Slack, and posted an expansion summary to Slack, so one of the six accounts went to a person based on the state of the case. The Incident Orchestrator routes incidents, enforces escalation timers and tracks resolution, which means making a routing decision, holding a timer across a gap the agent does not control, and following an incident through to resolution.

State persists across runs, agents run 24/7 in the background without anyone stepping in, and Coworker describes orchestrating several agents from a single control plane. A workflow involves the agent, the approver at a gate, the AE an account escalates to, and the connected systems each step reads and writes. Agents carry work through to the end, updating the CRM, creating tickets, sending follow-ups and drafting reports. Coworker does not describe retry, failure handling or what an agent does when a step fails.

Sourcecoworker.ai/agents capabilities, agent examples and FAQ, with /platformread 2026-09-12

A visual workflow builder offers built in control blocks including Router, Condition, Loop, Parallel and a Workflow block for composing sub workflows. Agents reason inside Agent blocks and act through tools, and Mothership, the natural language control plane, scaffolds and edits workflows, tables and knowledge bases across the workspace, with API and SDK build paths running alongside the canvas.

Sourcedocs.sim.ai/introductionread 2026-09-29

Agents start on a schedule, on real time events or manually. Scheduled triggers run agents for daily briefings, weekly reports and hourly monitoring. Custom triggers start agents on new Slack messages and Jira updates, CRM changes and calendar events, four event sources in four systems. Agents work where the team already works, running in Slack, joining meetings and updating boards instead of living in another tab.

The Coworker Native Apps (Work, Chat, Build, Meetings, Agents, Artifacts) are Coworker's own surfaces, and Coworker MCP puts the same agents inside Claude, Cursor, ChatGPT and Windsurf. Email appears in the agent examples, which draft outreach through Gmail. Agents run 24/7 in the background without anyone stepping in and send a notice when done, and none of this is held to one tier. Coworker does not say whether a customer can define a compound or conditional trigger beyond the four named event types.

Sourcecoworker.ai/agents capabilities and FAQread 2026-09-12

Six core triggers, Start, Schedule, Webhook, RSS Feed, Table and Sim Workspace Events, start workflows, and deployments expose any workflow to the outside world as an API, a chat interface or an MCP server.

Sourcedocs.sim.ai/introductionread 2026-09-29

OM2 reads every connected tool continuously, with no uploads or re-syncing, learns backwards over historical data at setup and keeps evolving as it sees more each day, whereas most assistants, Coworker says, fetch document chunks at query time and forget afterward. Knowledge is stored as connected facts about people, projects, customers and decisions, with customers, projects and deals, as well as people, as first class nodes rather than raw text. Recall is synthesized ahead of time rather than found at query time, relevancy-ranked by proximity, recency and centrality, with sub-second graph traversal.

Retrieval is scoped by permission-aware traversal built into the graph, with access policies on every fact and connection. Every answer traces back to its source, and agents get permissioned context across Slack, CRM, docs, Jira and 50+ apps. The corpus covers messages, tickets, deals, docs, sheets, calendars, CRM, emails, meetings, transcripts, code, wikis, boards, PRs, recordings and contacts. Coworker's comparative figures (84% preference, 20% faster, 66% cheaper) are its own unaudited benchmarks.

Sourcecoworker.ai/platform OM2 section with /blog/introducing-om2 and /agentsread 2026-09-12

Knowledge bases are a workspace resource of their own, with connectors, tag based filtering, chunking strategies and retrieval debugging, and agents retrieve relevant passages from customer documents to ground answers.

Sourcedocs.sim.ai/tools/knowledgeread 2026-09-29

OM2, which replaced OM1 as the headline layer, is built by reading the customer's connected tools continuously, with no uploads or re-syncing, and learns backwards over historical data at setup. Agents use the graph to track progress, learn from corrections and continuously improve, and memory compounds with every interaction, so learning from corrections is written by the agent as well as absorbed into the product.

Scoping rests on isolated single tenant infrastructure, access policies on every fact and connection, and permission aware traversal, with permissions per user in the native apps. Coworker states no retention period, expiry, purge or deletion path for the memory. The only published retention figure is 90 days for recorded meetings on the Enterprise tier.

Coworker MCP is one connection, read-only by default, so the graph can be read and inspected there and in the native apps, with permissions on each fact, but nothing lets a customer or an outside caller edit, correct or delete a fact in it.

Sourcecoworker.ai/platform, /agents and /blog/introducing-om2read 2026-09-12

Memory on the Agent block needs a conversation ID to persist across runs, and the agent can be set to one of four memory modes. None sends independent requests, Conversation holds the full history keyed by a conversation ID, Sliding Window by messages keeps the N most recent, and Sliding Window by tokens keeps messages up to a token limit.

The Mem0 integration lets agents keep persistent memory across workflow executions, recalling past conversations, remembering user preferences and building on previous interactions, with add, semantic search and retrieve operations, and Zep ships as a second external memory option. Tables hold durable structured rows that workflows read and write, and Variables carry in-run state.

Sourcedocs.sim.ai agent block, Mem0 tool and workflow documentationread 2026-08-31

Customers can add approval steps to any agent workflow, where a reviewer can review, edit or reject the agent's proposed action before it executes, so the hold sits on the agent's own action and the reviewer can amend it as well as approve or deny it. Both modes are supported, with agents running fully autonomously for routine tasks or with approval gates on any workflow so a person reviews and approves actions before they execute. Background runs work the same way, notifying the team when done so they can review the output and approve actions before they go live.

Permissions, approvals, credit limits and escalation logic are named together as controls, and a credit limit is a hard spending stop on an autonomous agent that does not depend on anyone noticing. Escalation logic appears in the products too, with the Incident Orchestrator enforcing escalation timers and the account agent escalating one strategic account to an AE. Agents inherit the permissions already in place, with no new logins and no new SSO maps, so an agent cannot go beyond the authority of the person it acts for.

Sourcecoworker.ai/agents capabilities and FAQ, with /platformread 2026-09-12

Human in the Loop and Guardrails ship as core workflow blocks alongside Condition and Wait, and enterprise access control enforces which blocks and model providers each permission group can use, both in the UI and at execution time.

Sourcedocs.sim.ai/introductionread 2026-09-29

Coworker holds SOC 2 Type II, GDPR compliance and CASA Tier 2, shown as three badges on the platform page that each link to a Secureframe trust portal at village-labs.secureframetrust.com. Identity and access controls include SSO, RBAC, and encryption at rest and in transit, and SSO is part of the Enterprise entitlements.

Permissions reach down to each fact, since Coworker is permission-aware by construction, with access policies on every fact and connection, so everyone sees only what they are allowed to see through traversal of the graph that respects permissions. Because the graph pulls Slack, CRM, email and HR systems together, that authorization on each fact governs what every user can reach. Agents inherit the permissions already in place, with no new logins and no new SSO maps, so there is no privilege elevation and no parallel access model to drift.

Isolated single tenant infrastructure and complete audit trails across the platform add to this, along with a commitment never to train models on customer data and published legal agreements, namely a privacy policy, terms and a Master Services Agreement.

The agents page says agents tap into 100k+ deployed agents' learnings, which reads as learning across customers and sits against the statement that customer data is never used to train models, and Coworker does not say whether the shared learnings are behavioral patterns rather than customer content.

Sourcecoworker.ai/platform enterprise-grade section and FAQ, with /pricingread 2026-09-12

Sim's Trust Center at trust.sim.ai lists SOC 2 Type II, ISO 27001 and GDPR, and SOC 2 compliance is among the Enterprise features. Access controls include SAML 2.0 and OIDC SSO across Okta, Entra ID, Google Workspace and ADFS, workspace roles and permission groups enforced at execution time, verified domains, session policies with maximum lifetime, idle timeout and sign out of all members, dedicated outbound IPs, secrets management and configurable data retention.

Sourcesim.ai homepage and pricing page, docs.sim.ai enterprise Security pageread 2026-09-29

Every action is logged and every decision auditable, with guardrails the customer sets and audit trails the customer owns, and the agents FAQ says every action in autonomous runs is logged for full auditability. Complete audit trails across the platform are listed among the enterprise controls beside permissions and identity, and logging every action records what the agent did, not only what it produced.

Answers are attributed separately in OM2, where every answer traces back to its source, so a conclusion can be walked back to the fact and the connected tool it came from. Coworker gives no detail on the audit trail itself, with no stated retention period and no word on whether it can be exported or filtered, whether it captures reasoning as well as actions, or whether it exists below the Enterprise tier.

Sourcecoworker.ai/agents and /platform enterprise-grade section, with /blog/introducing-om2read 2026-09-12

Logs record every run block by block as a workspace resource of their own, with logging and alerts, cost visibility for each run, enterprise audit logs tracking configuration and security relevant actions, and data drains that stream workflow logs, audit logs and Mothership data to a customer owned S3 bucket or HTTPS webhook on a schedule.

Sourcedocs.sim.ai/introductionread 2026-09-29

Isolated single tenant infrastructure is offered in four forms, cloud, private cloud, on premises or air gapped, and on premises and air gapped installs sit in the customer's own environment, so the customer decides where data rests. Coworker is SOC 2 Type II, GDPR and CASA Tier 2 verified, with SSO, RBAC, encryption, audit trails and air-gapped deployment support.

Single tenancy is the baseline, and the OM2 memory runs on isolated single-tenant infrastructure in Coworker's cloud or via VPC. VPC peering and bring your own model on the customer's own infrastructure are further options, so a customer who cannot send data to a shared service has three published routes.

The homepage says routing is all hosted in the US and the OM2 announcement leads with cloud and VPC, while the platform page states the full range. There is no named region list for the hosted option, no residency selection for cloud customers, and no page on what an air gapped deployment costs, excludes or requires.

Routing across twelve providers is hard to square with an air gapped install, and bringing your own model on the customer's infrastructure may resolve that, though Coworker does not say so.

Sourcecoworker.ai/platform OM2 security line and FAQread 2026-09-12

Sim runs cloud hosted at sim.ai or self hosted through Docker Compose and Kubernetes, including a fully offline setup with local models through Ollama, and self hosted deployments unlock the entire enterprise feature set through environment configuration, so the customer can control hosting entirely.

Sourcedocs.sim.ai/introductionread 2026-09-29

A catalog of agents is published, each described and tagged by role. Expansion Pipeline, for Account Management, finds accounts hitting usage ceilings and generates upgrade outreach, and Revenue Recovery, for Finance, processes payment failures with segment-specific recovery playbooks. Pipeline Cleanup, for Sales Ops, enforces CRM discipline on stale deals, past due dates and missing data.

Incident Orchestrator, for Engineering, routes incidents, enforces escalation timers and tracks resolution, and Data Analyst, for BI and Analytics, queries data, builds reports and surfaces insights automatically. Three more are named in the deployment section, Post-Meeting Actions for Sales, Contract Review for Legal and Bug Triage for Engineering, and the agent picker offers Account, MCP, Engineering, BI, Legal, Meetings, Finance and Sales as classes. Each agent works on its own and serves a different function on different objects for different teams in the same company.

Once a tool is connected, the product learns in minutes and suggests agents suited to the user's role, so the catalog proposes agents from what a customer connects. OM2 carries shared skills and artifact templates that travel across models and do work when selected, and use case pages by function, one each for Sales, Engineering, Customer Success, Support, Product, Marketing, People, Operations and Industries, make the catalog browsable. The named agents each appear as a card with no page of their own, and no count of the full library is given.

Sourcecoworker.ai/agents agent set and role suggestions, with /platform OM2 skillsread 2026-09-12

Agent skills provide reusable packaged capabilities and Mothership scaffolds complete systems from a natural language description, and Sim does not describe a template gallery, prebuilt agent library or marketplace.

Sourcedocs.sim.ai/introductionread 2026-09-29

Coworker routes each task to the optimal model across closed, open-source and self-hosted providers, balancing cost, latency and accuracy, with every task routed to the cheapest model that meets the quality bar. It names 12+ model providers, among them OpenAI, Anthropic and Google, plus open source Llama, Mistral and MiniMax. With BYOM, a customer can bring its own models and run them on its own infrastructure.

As better models ship, teams get them automatically with no migration and no vendor lock-in, and OM2's memory and skills travel with the customer whatever model it chooses. Coworker MCP, on the same page, lets the customer's own assistant reach Coworker's memory. Coworker does not say whether an administrator can pin a specific model per agent or can only rely on automatic routing.

Sourcecoworker.ai/platform routing section and FAQread 2026-09-12

More than fifteen providers are supported, including OpenAI, Anthropic, Google Gemini, Groq, Cerebras, DeepSeek, Mistral, xAI and OpenRouter, with bring your own key at base provider pricing and no markup, and local models run through Ollama or VLLM at no API cost. Enterprise permission groups can restrict which providers are allowed.

Sourcedocs.sim.ai/blocks/agentread 2026-09-29

Coworker MCP, a product with its own page, puts org memory, 50+ tools and dedicated agents inside any MCP client through one connection that is read-only by default and uses the same SSO and permissions, with Claude, Cursor, ChatGPT and Windsurf named as clients. OM2 also connects through MCP or an API, so Coworker's memory can be used outside Coworker in any MCP-compatible client.

Through it the customer's assistant calls Coworker, and because dedicated agents sit inside any MCP client, an outside assistant can invoke a Coworker agent that updates CRM records, creates tickets and sends follow ups. The agent picker lists MCP as one of its surfaces alongside Account, Engineering, BI, Legal, Meetings, Finance and Sales.

The connection reads by default, and the wording implies write access can be configured, but Coworker does not say how or what a write scope covers. Permissions and SSO come from the same identity, so the surface does not widen a customer's exposure. The memory is sold as one memory for every surface, with skills traveling too. Coworker publishes no tool list, endpoint reference, SDK or OpenAPI document, and the API named in the OM2 post has no published reference.

Sourcecoworker.ai/platform portability section and FAQ, with /blog/introducing-om2read 2026-09-12

Sim is open source under Apache 2.0 with the full source on GitHub, a public API and SDKs, a Function block for custom JavaScript, custom tools, and MCP in both directions, so agents consume MCP tools and any workflow deploys as an MCP server, with enterprise custom blocks on top.

Sourcedocs.sim.ai/introductionread 2026-09-29

There is no sandbox, dry run, simulation or staging workspace, no holdout, A/B comparison or versioning and rollback of an agent configuration, and no evaluator, scored result or regression surface. In the deployment lifecycle the agents FAQ walks through, a customer builds an agent in the Agent Builder and it runs, with no test step. The platform is organized around five layers, OM2, Routing, Portability, Integrations and Enterprise Grade, and none of them covers evaluation.

The Optimize and scale step says memory compounds with every interaction, which describes the memory layer growing. Agents are said to get smarter with use, observe the team's patterns and suggest new automations, and get smarter every week as they learn how the team works. Approval gates let a person review, edit or reject each action before it executes.

Sourcecoworker.ai/agents and /platformread 2026-09-12

The Evaluator block is a core block that uses AI to score and assess content quality against metrics the customer defines, each with a name, description and numeric range, with accuracy, clarity and relevance on one to five scales as worked examples.

The block generates a JSON Schema response format from those metrics and enforces strict mode so the model returns only numeric scores, and it supports quality control and A/B testing, including comparing several AI generated responses and assessing Agent block outputs to create feedback loops. A published pattern chains Agent to Evaluator to Condition against a threshold.

Guardrails, knowledge base retrieval debugging, block-by-block run traces and per-run cost visibility help with debugging. Sim does not describe keeping test sets, stored expected outputs or a regression history across versions.

Sourcedocs.sim.ai evaluator block and core blocks documentationread 2026-08-31

Data comes in through OAuth connectors feeding the graph, actions go out as connector writes that update CRM, create tickets and send follow-ups, and the outside surface is an MCP server. Coworker presents OM2 as replacing work at the interface, using sub-second graph traversal in place of brute force tool calling. Nothing operates a screen, and all fifty connectors are OAuth APIs.

Chrome and VS Code appear among the OAuth connectors on the platform page, in the list of connectors feeding the graph alongside Slack, Snowflake and Workday, and what is read from them includes code, PRs and recordings, all ingested into the knowledge graph. Coworker has no hosted or local browser, desktop session or remote computer control, and no RPA, recorder or extension that acts on a page.

Sourcecoworker.ai/platform integrations list and /agents context listread 2026-09-12

Browser Use and Stagehand ship as workflow tools in the integrations catalog alongside Apify and Bright Data, letting workflows drive real browsers against interfaces that expose no programmatic API, with Stagehand performing natural language navigation and action on live pages. These are third party browser engines wired in as catalog integrations, not a hosted or local browser or desktop session Sim runs itself.

Sourcedocs.sim.ai tools catalogread 2026-08-31