CAPABILITIES
Everywhere your work happens.
Agents that ask first, workflows you can watch, and your own branded app. Here's what each part does, how you use it, and how it compares.
Book a free call Watch the tour · 3:30
- 01 Team chat
- 02 Customers
- 03 The phone
- 04 Terminal
- 05 Your code
- 06 Browser
- 07 Overnight
- 08 Your app
THE TOUR · 3:30
See all eight stops in 3½ minutes.
Eight stops, narrated and captioned.
Everywhere your work happens
Same film as the homepage. A narrated tour of MyndHyve in eight stops. Team chat: mention your agent in Slack, Teams, Discord or Telegram and it answers in the thread, with the approval card right there. Customers: WhatsApp, SMS, email and Apple Messages for Business, with opt-outs honoured and anything unusual held for you. The phone: your agent answers a number, books the meeting and hands you the call. The terminal: Claude Code paired with your workspace, its prompts becoming approvals on your phone. Your code: TypeScript, Python and Go libraries, an API, signed webhooks and an MCP server. The browser: your agent fills in a form and stops before Submit. While you sleep: scheduled agents with memory, each job fired once. Your own app: web, desktop and phone, in your brand. Underneath, every step is on the record and agents never see your keys. It compares open-source agents, big-lab agents and MyndHyve.
THE PROBLEM
Your agents are scattered, and none of them ask first.
They live everywhere
One agent in a terminal, one in a chat app, one in a script nobody remembers writing. Nobody can say what's running.
They act without asking
A reply goes out with a discount you never offered. The tool says the task completed, and that's all it says.
They aren't yours
They run on someone else's product, under someone else's name, on models you didn't choose.
Eight stops
One agent. Every place your work already is.
STOP 1 · TEAM CHAT
Your agent, in the thread.
Mention your agent where your team already talks, and it answers in the thread. Every step that matters waits for a person.
The problem it solves Nothing customer-facing goes out because an agent decided it should.
- Start work from Slack — Available now
- Approvals in Teams, email, push — Available now
- Two-way chat in Slack, Teams, Discord, Telegram — Rolling out
How you use it
Connect Slack, Microsoft Teams, Discord or Telegram under Connections and add your agent to a channel. Mention it to ask a question or start a workflow. When it needs a decision, the approval card lands in the thread, or reaches you by email, in the browser or as a push notification. Approve, reject or request changes in one tap.
STOP 2 · YOUR CUSTOMERS
Customers text. You stay in charge.
Your agents answer customers on WhatsApp, by text message, by email and on Apple Messages for Business, in your voice and inside rules you set.
The problem it solves Fast replies at 2 am without a message you'd have to apologise for at 9.
- SMS — Available now
- Email — Available now
- WhatsApp Business — Rolling out
- Apple Messages for Business — Rolling out
How you use it
Connect a Twilio number for SMS, SendGrid, Postmark or your own SMTP server for email, and a WhatsApp Business sender (Twilio or Meta Cloud API). STOP and opt-out replies are honoured automatically and kept on a suppression list. Sends are recorded once, so a retry never messages the same person twice. Anything unusual, like an offer you never approved, waits for you.
STOP 3 · THE PHONE
Give your agent a number.
Your agent answers calls on its own number, talks like a person, books the meeting, and hands the call to you when it should.
The problem it solves No missed calls after hours, and no caller stuck with a bot when they need a person.
- In-app voice conversations — Available now
- Phone calls — Rolling out
How you use it
Add a phone number under Connections and choose which agent answers it, what it may book, and when it transfers to you. Every call is on the record with its transcript.
STOP 4 · THE TERMINAL
Everything has a command line.
Every workflow, run, toggle, key and webhook can be driven from the terminal. Claude Code can be paired with your workspace.
The problem it solves Coding agents that keep working when you step away, without giving them free rein.
- CLI — Available now
- Claude Code pairing — Available now
How you use it
Install the CLI with
npm i -g @openwop/cliand sign in with a device code. To pair Claude Code, switch on the Developer features, type the pairing code on your machine, and your ownclaudeshows up as a terminal tab in the app. Its permission prompts become approval cards you answer from anywhere. The machine only makes outbound connections, so no port is opened.STOP 5 · YOUR CODE
Plug it into your code.
Everything you can do in the app, your own software can do too, and other agents can call your workflows like tools.
The problem it solves You don't have to choose between a product your team can use and an engine your developers can build on.
- Libraries — Available now
- API — Available now
- Webhooks — Available now
- MCP server — Available now
- Agent-to-agent — Available now
How you use it
Use the libraries for TypeScript, Python or Go, or call the REST API with a workspace API key. Subscribe to signed webhooks to hear when runs finish, and trigger runs from inbound webhooks. Turn on the MCP server and Claude, ChatGPT and other MCP clients can call your workflows; the agent-to-agent endpoint lets other agents hand yours a task. You can also import skills in the open SKILL.md format into your signed pack catalog.
The libraries, API, webhooks, MCP server and agent-to-agent endpoint are all available now. Callers must sign in, each is bound to its own workspace, and run-starting calls are rate-limited.
STOP 6 · THE BROWSER
It fills the form. You press Submit.
Your agent opens a site, fills in the form and checks every detail. Then it stops right before Submit and waits for you.
The problem it solves The busywork of web forms, without an agent clicking Buy on your behalf.
- Browser control — Rolling out
How you use it
Give an agent a task and the sites it may visit. Reading and typing happen on their own; anything that commits, like Submit, Buy or Send, asks you first, and each session has a daily budget.
STOP 7 · WHILE YOU SLEEP
It works while you sleep.
Agents run on a schedule, pick up work from your board, and remember what matters.
The problem it solves No double emails, no double charges, and no work that silently stopped on Thursday.
- Schedules — Available now
- Board pickup — Available now
- Shared memory — Available now
How you use it
Give an agent a schedule or a recurring chat, or put cards on its board and let it take the next one. Memory is shared between people and agents. Automatic memory from chat is opt-in, and it only keeps what you actually said. Each scheduled slot fires once, even when several servers are running.
STOP 8 · YOUR OWN APP
Your own app. Everywhere.
MyndHyve prints the whole thing as your own branded app: logins, billing, a brand, a live site, and your agents, on the web, the desktop and your customers' phones.
The problem it solves The months between "the app works" and "the business works".
- Web app — Available now
- Signed desktop app — Rolling out
- iOS app — Rolling out
How you use it
Add your name, colours and domain, then switch on only the features you need from the feature toggles. Run it on MyndHyve, or self-host it as one container on Docker Compose, Fly, Render, AWS, Azure or Google Cloud, with SQLite or Postgres and any OpenID Connect sign-in.
UNDERNEATH EVERY STOP
Every step on the record.
Replay and rewind
Every run keeps an event log. Replay a run, or fork it from any step and try again.
Keys stay in the vault
Credentials are encrypted on the host and brokered per call. Agents never see your passwords or API keys.
Roles and single sign-on
Organisations with roles, SAML single sign-on and SCIM provisioning.
Signed packs
Add-on packs are signed with Ed25519 and checked against a trust store before they load.
WHO IT'S FOR
Same engine. Your brand.
Agencies
Print an app for every client, with your name on it. Their approvals, their data, your workflows. For agencies
Small shops
Answer customers overnight by text and email, with anything unusual held for the morning.
Founders
Ship the launch: emails, pages and follow-ups drafted from your own docs and sent when you say so. Campaign Studio
HOW IT COMPARES
Reach and control. Under your name.
| MyndHyve | OpenClaw | Hermes Agent | Claude Code + Cowork | OpenAI Codex + dots | Lindy | Manus | n8n | |
|---|---|---|---|---|---|---|---|---|
| Agent in your chat apps | ~ | ✓ | ✓ | ~ | ~ | ✓ | ~ | ✓ |
| Command line | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | ~ |
| SDKs and API | ✓ | ✓ | ~ | ✓ | ✓ | unverified | unverified | ✓ |
| MCP | ✓ | ✓ | ✓ | ✓ | ✓ | unverified | unverified | ✓ |
| Skills or plugin catalog | ~ | ✓ | ~ | ✓ | ✓ | ✗ | ~ | ~ |
| Memory | ✓ | ✓ | ✓ | ✓ | ~ | ✓ | unverified | ~ |
| Schedules and background work | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | unverified | ✓ |
| Browser or computer control | ✗ | ✓ | ✓ | ✓ | ✓ | ~ | ✓ | ✗ |
| Self-host | ✓ | ✓ | ✓ | ~ | ✗ | ✗ | ✗ | ✓ |
| Teams, roles and single sign-on | ✓ | ✗ | ✗ | ✓ | ✓ | ~ | unverified | unverified |
| Approval before an agent acts | ✓ | ~ | ~ | ✓ | ✓ | unverified | unverified | ~ |
| Audit trail and replay | ✓ | ✗ | ✗ | ✓ | ✓ | ✓ | unverified | unverified |
| Your brand on the app (white-label) | ✓ | ~ | ~ | ✗ | ✗ | ✗ | ✗ | unverified |
✓ ships · ~ partial or in preview · ✗ not offered · unverified: not confirmed from a primary source
MyndHyve marks what is in production on November 3, 2026, not what is rolling out; partial means available for some channels or behind a setting. OpenClaw and Hermes Agent are MIT-licensed, so "your brand" means forking them yourself. Sources: github.com/openclaw/openclaw and docs.openclaw.ai; github.com/NousResearch/hermes-agent; code.claude.com/docs/en/channels, platform.claude.com/docs/en/managed-agents/overview and support.claude.com Cowork release notes; learn.chatgpt.com/docs (Codex) and OpenAI's dots announcement (Sep 29, 2026); lindy.ai/pricing; manus.im; n8n.io and docs.n8n.io. Corrections welcome.
QUESTIONS
Know what you're choosing.
Which AI models does it use?
The one you choose. Agents run on Anthropic, OpenAI-compatible and other providers, and you can bring your own key.
Can an agent send something without me?
Only if you allow it. Every send, post and payment can require approval, and many do by default. Every step that happened is on the record.
Can we run it ourselves?
Yes. It ships as one container with deploy packs for Docker Compose, Fly, Render, AWS, Azure and Google Cloud.
How is this different from an open-source personal agent?
Those are built for one person on one machine, and they reach a lot of chat apps. MyndHyve is built for a business: many people, roles, approvals, a record of every step, and your own brand on the app.
What isn't ready yet?
Each stop above says what's available now and what's rolling out. Rolling out today: two-way chat in Slack, Teams, Discord and Telegram; WhatsApp Business and Apple Messages for Business conversations; phone calls; browser control; and signed desktop and iOS apps. The comparison marks only what is in production.
Print your business.
Free 15-minute call. We'll map your first agent, its workflow and its approvals.