# HexaHQ > HexaHQ turns what people teach and build with AI into reusable company knowledge and > workflows, shares them with the right people, and runs them against real business > systems under each organization's own approval rules. HexaHQ is a workflow-intelligence runtime that AI clients connect to over MCP. A person does useful work with their AI, keeps what worked as knowledge or a workflow, chooses who it is shared with (themselves, a team, the whole organization, or a named external collaborator), and other people reach the same material from whichever supported AI client they use. HexaHQ holds the credentials for connected systems, so an action runs as the person who approved it and leaves an audit record. HexaHQ works with tools that already ship an MCP server, and also with the larger set that have none: describe the API and the AI authors the capability, with no server to build or host. ## Start here - [One shared context for people and AI to get work done together](https://www.hexahq.ai/): HexaHQ automatically captures how your organization works, learns from corrections, and gives AI the context it needs to make decisions and take action the way your team would. - [Teach AI how your organization works as you work](https://www.hexahq.ai/product.html): HexaHQ captures the facts, rules, preferences, processes, and corrections that matter as people work, keeps them as one evolving source of truth, and gives each AI task the relevant context before it decides or acts. - [How HexaHQ gives your AI the context your company runs on](https://www.hexahq.ai/how-it-works.html): Start with a real task in the AI you already use. HexaHQ picks up the facts and rules that mattered, applies them to the work in front of you, and has them ready the next time they apply. - [Pricing](https://www.hexahq.ai/pricing.html): Start free: teach your AI how you work, connect your apps, and keep what it learns. Enterprise makes that context shared across the company and adds org-wide rules, the Hub, and a full record. Questions? sales@hexahq.ai. - [HexaHQ FAQ: business and technical questions](https://www.hexahq.ai/faq.html): Answers to what teams actually ask about HexaHQ: getting real work done across your apps, sharing with non-technical teammates, connecting apps with no MCP server, keeping your logins out of the chat, and how it compares to Claude Teams, ChatGPT, and MCP. ## What it connects to - [Connect any tool to your AI, no server to build](https://www.hexahq.ai/connect.html): Give your AI the apps you use in one click, even the ones with no MCP server. Popular apps connect directly; for everything else, HexaHQ builds the connection from the app's API, so you build and host nothing and your logins stay safe. - [Connect your AI to any app with an API](https://www.hexahq.ai/connect-any-api.html): Give your AI access to any app with an API. No MCP server required. HexaHQ runs tools that ship an MCP server, turns ordinary APIs into tools your AI can use, and supplies the company rules that should shape the action before it runs. - [Your company's context should not belong to one AI](https://www.hexahq.ai/multi-ai-workflows.html): HexaHQ keeps durable organizational context outside any one AI product, then makes the relevant parts available when you work in ChatGPT, Codex, Claude Chat, Claude Code, or Claude Cowork. - [Connect Figma to your AI, no server](https://www.hexahq.ai/connect/figma.html): Give your AI access to Figma in one click, or turn the Figma API into something your AI can use, with no server to build. Run Figma work as yourself, with your approval, and a full record. - [Connect GitHub to your AI, no server](https://www.hexahq.ai/connect/github.html): Give your AI access to GitHub in one click, or turn the GitHub API into something your AI can use, with no server to build. Run GitHub work as yourself, with your approval, and a full record. - [Connect Gmail to your AI, no server](https://www.hexahq.ai/connect/gmail.html): Give your AI access to Gmail in one click, or turn the Gmail API into something your AI can use, with no server to build. Run Gmail work as yourself, with your approval, and a full record. - [Connect Google Calendar to your AI, no server](https://www.hexahq.ai/connect/google-calendar.html): Give your AI access to Google Calendar in one click, or turn the Google Calendar API into something your AI can use, with no server to build. Run Google Calendar work as yourself, with your approval, and a full record. - [Connect Google Sheets to your AI, no server](https://www.hexahq.ai/connect/google-sheets.html): Give your AI access to Google Sheets in one click, or turn the Google Sheets API into something your AI can use, with no server to build. Run Google Sheets work as yourself, with your approval, and a full record. - [Connect HubSpot to your AI, no server](https://www.hexahq.ai/connect/hubspot.html): Give your AI access to HubSpot in one click, or turn the HubSpot API into something your AI can use, with no server to build. Run HubSpot work as yourself, with your approval, and a full record. - [Connect Intercom to your AI, no server](https://www.hexahq.ai/connect/intercom.html): Give your AI access to Intercom in one click, or turn the Intercom API into something your AI can use, with no server to build. Run Intercom work as yourself, with your approval, and a full record. - [Connect Jira to your AI, no server](https://www.hexahq.ai/connect/jira.html): Give your AI access to Jira in one click, or turn the Jira API into something your AI can use, with no server to build. Run Jira work as yourself, with your approval, and a full record. - [Connect Linear to your AI, no server](https://www.hexahq.ai/connect/linear.html): Give your AI access to Linear in one click, or turn the Linear API into something your AI can use, with no server to build. Run Linear work as yourself, with your approval, and a full record. - [Connect Notion to your AI, no server](https://www.hexahq.ai/connect/notion.html): Give your AI access to Notion in one click, or turn the Notion API into something your AI can use, with no server to build. Run Notion work as yourself, with your approval, and a full record. - [Connect QuickBooks to your AI, no server](https://www.hexahq.ai/connect/quickbooks.html): Give your AI access to QuickBooks in one click, or turn the QuickBooks API into something your AI can use, with no server to build. Run QuickBooks work as yourself, with your approval, and a full record. - [Connect Salesforce to your AI, no server](https://www.hexahq.ai/connect/salesforce.html): Give your AI access to Salesforce in one click, or turn the Salesforce API into something your AI can use, with no server to build. Run Salesforce work as yourself, with your approval, and a full record. - [Connect Slack to your AI, no server](https://www.hexahq.ai/connect/slack.html): Give your AI access to Slack in one click, or turn the Slack API into something your AI can use, with no server to build. Run Slack work as yourself, with your approval, and a full record. - [Connect Snowflake to your AI, no server](https://www.hexahq.ai/connect/snowflake.html): Give your AI access to Snowflake in one click, or turn the Snowflake API into something your AI can use, with no server to build. Run Snowflake work as yourself, with your approval, and a full record. - [Connect Stripe to your AI, no server](https://www.hexahq.ai/connect/stripe.html): Give your AI access to Stripe in one click, or turn the Stripe API into something your AI can use, with no server to build. Run Stripe work as yourself, with your approval, and a full record. - [Connect Zendesk to your AI, no server](https://www.hexahq.ai/connect/zendesk.html): Give your AI access to Zendesk in one click, or turn the Zendesk API into something your AI can use, with no server to build. Run Zendesk work as yourself, with your approval, and a full record. ## Use cases by team - [Data & Analytics use cases](https://www.hexahq.ai/use-cases/data.html): What data teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Engineering use cases](https://www.hexahq.ai/use-cases/engineering.html): What engineering teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Finance use cases](https://www.hexahq.ai/use-cases/finance.html): What finance teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Use cases by team](https://www.hexahq.ai/use-cases/): What every team gets done in one sentence with HexaHQ: sales, marketing, product, engineering, data, RevOps, finance, and leadership. Real, searchable examples per team, each reaching across the apps where your work lives. - [Leadership use cases](https://www.hexahq.ai/use-cases/leadership.html): What leadership teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Marketing use cases](https://www.hexahq.ai/use-cases/marketing.html): What marketing teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Product use cases](https://www.hexahq.ai/use-cases/product.html): What product teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [RevOps use cases](https://www.hexahq.ai/use-cases/revops.html): What RevOps teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. - [Sales use cases](https://www.hexahq.ai/use-cases/sales.html): What sales teams do in one sentence with HexaHQ: workflows that pull from your real tools and run end to end. Not one-off tasks; workflows HexaHQ learns and your whole team reuses. ## Answers to common questions - [Why AI keeps making the same mistake, and how to make corrections stick](https://www.hexahq.ai/answers/ai-learns-from-corrections/): A correction goes into the conversation, not into anything the next conversation reads. To make it stick it has to be saved with the scope it applies to, and read back on the next task where it matters. How that loop works, and where it stops. - [How to keep the same AI context when you switch between ChatGPT and Claude](https://www.hexahq.ai/answers/cross-model-context/): Your company's AI memory should belong to your company, not to one model. Keep the durable rules outside both products, and let each AI receive the part that applies to the task in front of it. - [Do I need to build my own AI memory system?](https://www.hexahq.ai/answers/diy-ai-memory-stack/): Markdown, Git, a vector database, a memory server, and a handoff file will beat most products for one person. Here is the specific point where the build turns into a product, and the five things that make it one. - [Answers: AI memory, shared context, and safe AI action](https://www.hexahq.ai/answers/): Direct answers to nine questions people ask about making AI useful at work, in two groups: keeping what the AI has been told, and making it get things right before it acts. Each page answers the question first and ends with the AI doing real work. - [How to give ChatGPT or Claude persistent memory across sessions](https://www.hexahq.ai/answers/persistent-ai-memory/): The model itself cannot remember, so persistent memory has to live outside it and be handed back on the next task. What people build today, where it holds up, and what changes once the memory has to be corrected, shared, and connected to real work. - [Why enterprise RAG returns outdated or conflicting answers](https://www.hexahq.ai/answers/rag-outdated-conflicting-answers/): Retrieval finds source material that looks relevant. Relevance says nothing about which statement is current or authoritative. The five questions an index cannot answer, why better search does not fix them, and what to keep instead. - [How to give AI role-based access to company knowledge](https://www.hexahq.ai/answers/role-based-ai-knowledge-access/): One company knowledge base is not enough if every person's AI receives all of it. How to keep one shared context layer while each person and their AI gets only the relevant, permitted parts. - [How to let AI take real actions in your systems safely](https://www.hexahq.ai/answers/safe-ai-actions/): Two things have to be true at once: the AI needs your company's rules for the situation before it decides, and the write has to run under a policy you set. Why read-only agents and broad service accounts both fall short. - [How to share AI context across a team](https://www.hexahq.ai/answers/shared-team-ai-context/): Shared projects and workspaces can centralize source material. Turning what one person teaches AI during real work into canonical organizational context, with an audience, an owner, and a place where it applies, is the harder part. - [Why AI agents go off track on long, multi-step work](https://www.hexahq.ai/answers/why-ai-agents-go-off-track/): One wrong assumption at step one is carried into every step after it. Why more autonomy and more supervision both miss, and what changes when the agent gets the company's relevant rules before the first consequential decision. ## Comparisons - [Do companies need AI employees, or AI extensions of their existing team?](https://www.hexahq.ai/compare/ai-employees.html): The AI employee is a good picture and a poor unit. What makes a colleague valuable is years of accumulated context. For most knowledge work the better shape is several specialized AI extensions of each person, each inheriting the right context and permissions. - [Can ChatGPT and Claude share the same memory?](https://www.hexahq.ai/compare/chatgpt-claude-memory.html): Not natively. Each vendor's memory belongs to one person and one vendor, and the imports on offer are hand-carried snapshots. Your company's AI memory should belong to your company, not to one model. What that means in practice, and who should receive which context. - [HexaHQ vs ChatGPT agents: governed action for your team](https://www.hexahq.ai/compare/chatgpt.html): ChatGPT's agent mode can run a task for one person, within monthly usage caps. HexaHQ turns your team's AI into governed action across your real business tools, with per-person credentials, an approval gate, and an audit trail, run through the AI client you already use. - [HexaHQ vs Claude Teams: who governs what the AI does](https://www.hexahq.ai/compare/claude-teams.html): Claude Teams gives shared billing and an org Capabilities console. HexaHQ governs what the AI actually does, per person, with a shared setup non-technical teammates can run, logins that never touch the chat, and an audit trail. A companion, not a replacement. - [Should I keep my AI memory in GitHub?](https://www.hexahq.ai/compare/github-repo-memory.html): For code conventions, yes. AGENTS.md and a repo of Markdown are a good home for anything an engineer maintains next to the code. Business context is a different job: the people who own it do not open pull requests, and the merge that matters is between meanings, not lines. - [Is Glean the context layer for enterprise AI?](https://www.hexahq.ai/compare/glean.html): Glean builds and selects enterprise context from connected systems and work signals. HexaHQ turns explicit teaching plus organizational evidence into recorded context about what AI should apply, including when the organization has not settled the answer. Where each one fits, checked against Glean's current platform. - [Why can't I just let my AI read our Google Drive?](https://www.hexahq.ai/compare/google-drive.html): You can, and you should keep Drive. ChatGPT and Claude both read it with your own permissions. The trouble starts when four documents state the same rule differently and the AI has to decide which one is current, where it applies, and whether you should see it. - [Compare HexaHQ](https://www.hexahq.ai/compare/): How HexaHQ compares to what teams already use for AI context and action: Glean, Google Drive, Obsidian, GitHub, native ChatGPT and Claude memory, Claude Teams, Manus, Zapier and n8n, and the Skills-vs-MCP question. - [HexaHQ vs Manus: for teams, after the demo ends](https://www.hexahq.ai/compare/manus.html): Manus can run recurring tasks for the person who set them up. HexaHQ runs them for your whole team, under one approval policy, with a shared credential model and an audit trail. What happens after the demo ends. - [Can I use Obsidian as long-term memory for Claude or ChatGPT?](https://www.hexahq.ai/compare/obsidian.html): Yes, and for one person it works well. A vault plus an MCP server gives your AI durable Markdown memory you own. Here is where a personal vault stops being enough for a company, and what HexaHQ does about audience, reconciliation, and action. - [Claude Skills vs MCP vs HexaHQ workflows](https://www.hexahq.ai/compare/skills-vs-mcp.html): Skills change how the AI reasons. MCP gives it access to tools. HexaHQ workflows govern what it is allowed to do in your systems, and prove it. The plain-language explainer with the governance axis every other piece leaves out. - [HexaHQ vs Zapier & n8n: connector custody for AI](https://www.hexahq.ai/compare/zapier-n8n.html): Zapier, Make, and n8n automate well. The question is not whether the run repeats. It is whose connectors your AI acts through, who holds the credentials, and how the AI decides what to run. HexaHQ turns any API into a governed tool your org controls. ## Blog - [Blog](https://www.hexahq.ai/blog/): Writing from HexaHQ on how AI actually does knowledge work: the patterns behind retrieval, organizational context, reasoning, and taking action in real business systems. - [ETL Built the Assembly Line. RARE Lets AI Drive.](https://www.hexahq.ai/blog/rare-retrieve-augment-reason-execute/): RARE is an AI integration pattern: Retrieve the live system state, Augment it with organizational knowledge, Reason with a model, Execute through APIs. Where ETL follows a route decided in advance, RARE lets the AI choose the route from live data and how your organization works. - [The Most Important Context Your AI Needs Was Never Written Down](https://www.hexahq.ai/blog/the-most-important-context-your-ai-needs-was-never-written-down/): Nobody writes a handbook page explaining how your company uses the Next Step field in Salesforce. Most operating context is tacit, so enterprise search cannot retrieve it. Why AI needs that context captured, inferred, and reconciled before it can act consistently. ## Help center - [Guides & comparisons](https://www.hexahq.ai/guides/): Plain-language answers on giving AI the context your company runs on: how AI learns how a company works, how a team corrects it once instead of in every chat, whether ChatGPT and Claude can use the same company context, Skills vs MCP vs workflows, and how HexaHQ compares to Claude Teams, ChatGPT, Manus, Zapier & n8n, Glean, Google Drive, Obsidian, and GitHub. - [Add HexaHQ to ChatGPT](https://www.hexahq.ai/help/add-hexa-chatgpt.html): Step by step: how to add the HexaHQ connector to ChatGPT on your own plan. Developer mode goes on under Security and login, then the plugin is created from the Plugins page with HexaHQ's MCP address and OAuth. - [Add HexaHQ to Claude Team or Enterprise](https://www.hexahq.ai/help/add-hexa-claude-team.html): Step by step for a Claude Team or Enterprise workspace: an Owner adds the HexaHQ connector once in Organization settings, and then each member connects their own account. Both halves, with every screen shown. - [Add HexaHQ to Claude on your own plan](https://www.hexahq.ai/help/add-hexa-claude.html): Step by step: how to add the HexaHQ connector to Claude on a Pro or Max plan. You add it yourself under Customize, and connecting uses OAuth in the browser rather than a pasted secret. - [Capabilities in HexaHQ](https://www.hexahq.ai/help/capabilities/): Learn how capabilities define the specific actions HexaHQ can take in a connected tool. - [How to check Google Search Console with ChatGPT](https://www.hexahq.ai/help/check-google-search-console-with-chatgpt/): Connect ChatGPT to your Google Search Console so it can tell you whether a page is in Google, when Google last crawled it, and how it is doing in search. Every screen shown. - [How to check Google Search Console with Claude](https://www.hexahq.ai/help/check-google-search-console-with-claude/): Connect Claude to your Google Search Console so it can tell you whether a page is in Google, when Google last crawled it, and how it is doing in search. Every screen shown. - [Collections in HexaHQ](https://www.hexahq.ai/help/collections/): Learn how collections control which members of an organization can see shared knowledge. - [Connectors in HexaHQ](https://www.hexahq.ai/help/connectors/): Learn how connectors give HexaHQ secure access to the apps and APIs where your work happens. - [Help center](https://www.hexahq.ai/help/): How to set up and use HexaHQ: connect your tools, turn any API into something your AI can call, run and share workflows, approvals and the audit trail, the Hub, and account and billing. Searchable, task-by-task. - [Knowledge in HexaHQ](https://www.hexahq.ai/help/knowledge/): Learn how HexaHQ stores, versions, applies, and shares the context your AI uses. - [HexaHQ app URLs for an OAuth developer console: host domain, launch, disconnect, connect](https://www.hexahq.ai/help/oauth-app-urls.html): The exact URLs to enter when you register a HexaHQ OAuth app in a developer console: host domain app.hexahq.ai, launch URL, disconnect URL, connect and reconnect URL, privacy policy, and terms. Covers the Intuit QuickBooks production form, Google, and Microsoft. - [Refresh the HexaHQ connector in ChatGPT](https://www.hexahq.ai/help/refresh-connector-chatgpt.html): Step by step: how to refresh the HexaHQ MCP connector in ChatGPT. ChatGPT caches HexaHQ's tool list per connector version, so a refresh needs a new chat as well, and a tool ChatGPT says it does not have is usually just unloaded rather than out of date. - [Refresh the HexaHQ connector in Claude Chat or Claude Cowork](https://www.hexahq.ai/help/refresh-connector-claude.html): Step by step: how to refresh the HexaHQ MCP connector in Claude Chat and Claude Cowork, and how to tell whether you need to. Claude re-reads HexaHQ's tool list every time it connects, so a new chat usually picks up a change on its own, and there is an explicit Refresh tools list control when it has not. - [Runs in HexaHQ](https://www.hexahq.ai/help/runs/): Learn how runs record the steps, approvals, and audit history for work carried out through HexaHQ. - [Sharing folders and files, inside and outside your company](https://www.hexahq.ai/help/sharing/): Who can share in HexaHQ, why you might not be able to, how to share a folder or file with someone outside your company, and what a share folder is. - [How to use ChatGPT with your WordPress site](https://www.hexahq.ai/help/use-chatgpt-with-wordpress/): Connect ChatGPT to your own WordPress site so it can write posts in your voice and save them as drafts, then publish only after you approve. Every screen shown, no plugin to install. - [How to use Claude with your WordPress site](https://www.hexahq.ai/help/use-claude-with-wordpress/): Connect Claude to your own WordPress site so it can write posts in your voice and save them as drafts, then publish only after you approve. Every screen shown, no plugin to install. - [Workflows in HexaHQ](https://www.hexahq.ai/help/workflows/): Learn how HexaHQ turns successful work into a repeatable process that your team can run again. - [Managing and deleting your data](https://www.hexahq.ai/help/your-data/): See what HexaHQ holds, disconnect a tool, remove a saved sign-in, revoke access at the source, and delete your data. ## Company and legal - [About HexaHQ: AI should learn how your organization works](https://www.hexahq.ai/about.html): Generic AI is useful and still does not know how your company works. HexaHQ exists so AI can take on parts of a person's job using the context that person and their organization actually rely on, and so that context belongs to the company. - [AI Addendum](https://www.hexahq.ai/ai-addendum.html): HexaHQ AI Addendum (Common Paper Standard AI Addendum v1.0). - [How does AI learn how your company actually works?](https://www.hexahq.ai/company-knowledge.html): HexaHQ captures useful context as people work, learns from corrections, keeps shared concepts aligned, and gives each AI task only the parts that apply, without asking your team to maintain a second documentation system. - [Contact Form](https://www.hexahq.ai/contact-form/): Contact HexaHQ. Email us and we will route your message to the right person. - [Contact HexaHQ](https://www.hexahq.ai/contact.html): Get in touch with HexaHQ: talk to sales about teams and enterprise, or start free and try it on your own tools. - [Data Processing Agreement](https://www.hexahq.ai/data-processing-addendum/): HexaHQ Data Processing Agreement (Common Paper DPA Standard Terms v1.1). - [Privacy Policy](https://www.hexahq.ai/privacy-policy/): HexaHQ Privacy Policy: how Salesbot Labs, Inc. (dba HexaHQ) collects, uses, and protects your information, including our Google API Limited Use commitment. - [Privacy Request](https://www.hexahq.ai/privacy-request/): Request access to, correction of, or deletion of the personal information HexaHQ holds about you. Email privacy@hexahq.ai to make a request. - [Security Policy](https://www.hexahq.ai/security-policy/): HexaHQ's Security Policy: how Salesbot Labs, Inc. (dba HexaHQ) protects customer data, covering governance, credential custody, encryption, access control, infrastructure, and incident response. - [Security & governance](https://www.hexahq.ai/security.html): How HexaHQ keeps your AI safe: your logins stay safe and never show up in the chat, everyone acts as themselves, you approve anything that sends or changes, and every action is recorded, so you can prove what the AI did and on whose say-so. - [Subprocessors](https://www.hexahq.ai/subprocessors.html): The third-party subprocessors HexaHQ uses to provide the service, what they process, and where. Connected providers and your AI client act on your instruction and are not HexaHQ subprocessors. - [Turn what your team knows into context the right people and their AI can use](https://www.hexahq.ai/teams.html): HexaHQ helps your team capture useful knowledge as work happens, keep shared concepts aligned, and make the right context available to the people who need it. Then identity, approval policy, and audit make it safe for AI to act. - [Terms of Service](https://www.hexahq.ai/terms-of-service/): HexaHQ Terms of Service.