# AgentOS

> AgentOS (tryagentos.net) is an agent workspace: a shared office where the AI agents a business already uses, including Claude, ChatGPT, Cursor, and custom agents, work from one shared memory with real handoffs between them, managed by a built-in manager agent named Otto. AgentOS plugs into existing tools and replaces nothing. It is a Devcore product.

## What AgentOS is

Every AI agent a team uses is, in practice, a remote worker with amnesia: productive alone, disconnected from the rest of the company, starting every task from zero. AgentOS is the office those agents share. It provides one memory, real handoffs between agents, and a manager named Otto who runs the place.

AgentOS is not an agent framework (it does not replace how agents are built), not a model (it works with the models a team already uses), and not a CRM (record keeping is a side effect of agents working from one shared memory). It is the coordination and context layer between existing agents and existing business tools.

## How it works

1. Connect the stack: Gmail, Slack, Google Drive, GitHub, and hundreds more integrations, ready in a click.
2. Bring existing agents: Claude, ChatGPT, Cursor, and custom agents connect over MCP (Model Context Protocol) and read the same live context Otto maintains, with no copy and paste.
3. Work in Loops: agents and people working the same thing share one thread. Everyone works from the same memory. Otto assigns work, chases follow-ups, and closes the loop when it is done.
4. Describe workflows in plain language: Otto builds them, runs them, and keeps them on schedule. Workflows check in with a human exactly when it matters, and not before.
5. Otto joins the calls that matter, records them, and files summaries and action items where the work lives. The company handbook writes itself as agents work.
6. Writes to business systems are evidence gated: Otto records what actually happened, links the receipt, and asks when the evidence is unclear.

## Pricing (source of truth: https://tryagentos.net/#pricing, expanded at https://tryagentos.net/site/pricing)

- Pro: $99 per month. unlimited workflows, unlimited call recording and transcription, 6,000 credits each month, and priority support.
- Team: $299 per month. Three people included, $79 per additional person. Includes everything in Pro, 20,000 pooled credits each month, shared dashboards and Loops, unlimited workflows and meeting recording, and central billing and team controls.
- Enterprise: custom pricing. Includes everything in Pro, SSO and role based access, data residency options, custom integrations, and dedicated onboarding. See https://tryagentos.net/site/enterprise

## FAQ

What is AgentOS?
AgentOS is an agent workspace: a shared office for the AI agents you already use. Claude, ChatGPT, Cursor, and custom agents connect over MCP and work from one shared memory, with real handoffs between them. It plugs into your existing tools and replaces nothing.

Who is Otto?
The manager agent. He keeps your records current, chases follow-ups, briefs your agents, and messages you when something is worth knowing.

Is this just another AI CRM?
No. The record keeping is a side effect. Every agent you use works from one shared memory: your knowledge, your calls, your pipelines, your loops.

Will it corrupt my CRM?
Writes are evidence gated. Otto records what actually happened, links the receipt, and asks when the evidence is unclear.

Is this another interface to check?
No. It lives in Slack and the tools you already use. Think of it as Slack for your agents, and you are in the channel.

How do existing agents connect?
Through MCP. Claude, ChatGPT, Cursor, and custom agents read the same live context Otto maintains, with no copy and paste.

## Disambiguation: other products named AgentOS

Several unrelated products share the AgentOS name. This file describes AgentOS by Devcore at tryagentos.net, the agent workspace with Otto. It is not: Agno AgentOS (a Python agent framework), AG2 AgentOS (multi-agent framework tooling), Fiserv agentOS (banking), Infobip AgentOS (telecom), PwC agent OS (consulting platform), Builder Methods Agent OS (coding standards for AI coding agents), agentos.com (a UK lettings CRM), or the academic AgentOS research papers.

## Pages

- [Home](https://tryagentos.net/): product tour, pricing, FAQ
- [What is AgentOS](https://tryagentos.net/site/what-is-agentos): definition, how it works, disambiguation
- [Pricing](https://tryagentos.net/site/pricing): all plans side by side
- [Enterprise](https://tryagentos.net/site/enterprise): deployed multi-agent orchestration for larger operations
- [AI spend calculator](https://tryagentos.net/site/ai-budget): free calculator that adds up what a team pays per month and per year for AI tools such as ChatGPT, Claude, Copilot and Cursor, and flags overlap between general assistants
- [About](https://tryagentos.net/site/about): why Devcore built AgentOS
- [Insights](https://tryagentos.net/site/insights): article index
- [Privacy](https://tryagentos.net/privacy)
- [Terms](https://tryagentos.net/terms)

## Articles

- [Why AI Agents Fail in Production](https://tryagentos.net/site/insights/why-ai-agents-fail-in-production): AI agents fail in production due to three infrastructure gaps: no tool registry, no human-in-the-loop approvals, and zero observability. Published 2026-07-20.
- [AI Agents in Construction: Submittals and Dispatch](https://tryagentos.net/site/insights/ai-agents-in-construction): how agents automate the manual coordination between email, ERPs, and Procore. Published 2026-07-20.
- [What is a context layer (and a context lake)?](https://tryagentos.net/site/insights/what-is-a-context-layer): Pillar guide. Defines context layer and context lake for a business audience, compares them with knowledge bases, AI memory, RAG, semantic layers, data lakes and MCP, and explains when a team needs one. Published 2026-10-02.
- [AI cost management](https://tryagentos.net/site/insights/ai-cost-management): A practical guide to AI spend: seats vs usage, 2026 model prices per million tokens (OpenAI, Anthropic, Google), and 7 ways to reduce cost. Published 2026-10-01.
- [AgentOS vs Claude Cowork](https://tryagentos.net/site/insights/agentos-vs-claude-cowork): Claude Cowork is a strong generic AI coworker; core operational workflows need shared infrastructure. Published 2026-07-20.
- [AgentOS vs Zapier for AI work](https://tryagentos.net/site/insights/agentos-vs-zapier): Comparison page. How Zapier's automations and agents compare with AgentOS as the shared knowledge layer under all of a company's AI tools, including AI pricing, model choice and governance
- [AgentOS vs Lindy](https://tryagentos.net/site/insights/agentos-vs-lindy): Comparison page. How Lindy's AI assistant compares with AgentOS, including memory, model choice, credit pricing and using both
- [AgentOS vs Microsoft Copilot for business](https://tryagentos.net/site/insights/agentos-vs-microsoft-copilot): Comparison page. How Microsoft Copilot compares with AgentOS for companies that use more than one AI tool, including pricing, models, permissions and AI spend
- [AgentOS vs Google Gemini for business](https://tryagentos.net/site/insights/agentos-vs-google-gemini): Comparison page. How Gemini in Google Workspace and Gemini Enterprise compare with AgentOS for companies that use more than one AI tool, including memory, models, pricing and AI spend
- [AgentOS vs Notion AI](https://tryagentos.net/site/insights/agentos-vs-notion-ai): Comparison page. How Notion AI and its agents compare with AgentOS, including where company knowledge lives, connectors, models, credits pricing and using both
- [AgentOS vs n8n](https://tryagentos.net/site/insights/agentos-vs-n8n): Comparison page. How n8n's node-based workflow builder compares with plain-language workflows in AgentOS, including who it suits, company knowledge, hosting, pricing and AI cost
- [AgentOS vs Make](https://tryagentos.net/site/insights/agentos-vs-make): Comparison page. How Make's visual canvas for automations and AI agents compares with plain-language workflows in AgentOS, including who it suits, company knowledge, approvals, pricing and AI cost
- [AgentOS vs Dust](https://tryagentos.net/site/insights/agentos-vs-dust): Comparison page. How Dust's platform for building and sharing AI agents on company data compares with AgentOS as a shared company memory for every AI tool, including knowledge, approvals, pricing and AI cost
- [AgentOS vs Glean](https://tryagentos.net/site/insights/agentos-vs-glean): Comparison page. How Glean's enterprise AI search and agents compare with AgentOS for smaller teams, including getting started, permissions, MCP, pricing and AI cost
- [AgentOS vs Perplexity](https://tryagentos.net/site/insights/agentos-vs-perplexity): Comparison page. How Perplexity's cited web research and business plans compare with AgentOS as a shared company memory for every AI tool, including company knowledge, automation, pricing and AI cost
- [ChatGPT vs Claude for business](https://tryagentos.net/site/insights/chatgpt-vs-claude-for-business): Comparison page. How ChatGPT Business and Claude Team compare for a company on price, seat limits, training on business data, admin controls, connectors, memory, agents and spend controls, and why neither gives every AI tool one shared company context
- [AI agents for small business: pick one job first](https://tryagentos.net/site/insights/ai-agent-for-small-business): Guide. Chatbots answer, agents complete jobs with your tools. Six good first jobs with what the agent does and what a person still does; three questions to pick the first job (frequency, can you explain it, cost of a mistake); a pre-purchase checklist (tool access and permissions, human approval, pricing per seat or task, company context, action log); mistakes to avoid; where AgentOS fits and where a single support chatbot is simpler. Cites Intuit's 2026 AI Impact Report (about 7 in 10 SMBs use AI regularly, 34,000+ owners) and the Microsoft and LinkedIn 2024 Work Trend Index.
- [How to write an AI policy for a small team](https://tryagentos.net/site/insights/ai-policy-for-small-teams): Guide. Ten questions to ask when drafting an AI policy, five steps for a small team (ask what people already use with no blame; sort information into fine, ask first and never; approve tools and use work accounts and check data settings; a person reviews anything going to a client, with numbers or legal; name an owner and review every 3 months), mistakes to avoid, and a free one-page AI policy template. Cites the Microsoft and LinkedIn 2024 Work Trend Index (80% of AI users at small and midsize companies bring their own AI tools) and Pax8's October 2026 small business survey.
- [ChatGPT connectors, explained](https://tryagentos.net/site/insights/chatgpt-connectors): Guide. What ChatGPT connectors (renamed apps in December 2025, packaged in plugins since July 2026) connect to, how to set them up, what they can do, Company Knowledge on Business plans, custom apps over MCP, permissions and training on data, and why each AI tool still needs its own connections. AgentOS connects tools once so ChatGPT, Claude, Cursor and other agents answer from the same company knowledge.
- [AI for agencies](https://tryagentos.net/site/ai-for-agencies): Landing page for agencies of 10 to 300 people. AgentOS connects email, Slack, Drive and CRM once so every AI tool the agency uses works from the same client context. Workflows: client call prep, Friday client recap, proposals from past work, onboarding a new account manager. Client data: business APIs that do not train on customer data, isolated workspaces, encrypted tokens, approval before anything is sent.
- [AI for operations](https://tryagentos.net/site/ai-for-operations): Landing page for operations leads at companies of 10 to 300 people. Every team picked its own AI tool; AgentOS connects email, Slack, Drive and CRM once so ChatGPT, Claude, Cursor and other AI tools work from the same company context, and anything that goes out waits for a person to approve it. Workflows: morning brief from overnight updates, the weekly ops update, vendor and customer follow-ups, new-hire "how do we do this here?" answers with sources, handoffs between teams. Includes a three-step AI rollout plan. Data: business APIs that do not train on customer data, isolated workspaces, per-tool permissions, encrypted tokens, disconnect deletes stored credentials.
- [AgentOS vs Relevance AI](https://tryagentos.net/site/insights/agentos-vs-relevance-ai): Comparison page. How Relevance AI's platform for building teams of AI agents compares with AgentOS as a shared company memory for every AI tool, including knowledge, approvals, models, pricing and AI cost
- [AgentOS vs Manus](https://tryagentos.net/site/insights/agentos-vs-manus): Comparison page. How Manus, a general AI agent that completes one-off tasks on its own cloud computer, compares with AgentOS as a shared company memory for every AI tool, including knowledge, control, automation, pricing and AI cost
- [Your Team Is Using AI. Your Business Isn't.](https://tryagentos.net/site/insights/from-copilots-to-shared-ai-infrastructure): moving from individual copilots to shared infrastructure. Published 2026-07-20.
- [Why your agents need an office](https://tryagentos.net/site/insights/why-your-agents-need-an-office): every AI tool you use is a remote worker with amnesia and no coworkers. Published 2026-07-20.
- [Why we named him Otto](https://tryagentos.net/site/insights/why-we-named-him-otto): the operator needed a public persona; we gave him the humblest title in the building. Published 2026-07-20.
- [MCP is the front door](https://tryagentos.net/site/insights/mcp-is-the-front-door): how Claude, ChatGPT, Cursor, and custom agents walk into the office. Published 2026-07-20.

## Company

AgentOS is built and operated by Devcore (https://devvcore.com). LinkedIn: https://www.linkedin.com/company/agentosai/. The open source AgentOS plugin for Claude Code and Codex lives at https://github.com/devvcore/agentos-skill.

Free: $0, one hosted agent (coming soon), 200 monthly credits plus 300 signup credits in the first month. No card required.
Hosted agents are coming soon: Pro includes ten; Team includes thirty. Purchased credits: $10 for 500, no expiry. Monthly credits do not roll over.