HeyBob vs Hermes: managed vs self-hosted
Matthew, founder of Bob · June 16, 2026
Key takeaways: Hermes (hermes-agent from Nous Research) is an open-source, self-hosted AI agent you run on your own infrastructure, anywhere from a $5 VPS to a GPU cluster. HeyBob is a managed AI employee: we run the runtime, Bob lives in Slack and Microsoft Teams, pauses for an Approve/Deny gate before writes, and ends every run with an itemized receipt. Same category as OpenClaw; the opposite operating model to Bob.
Two different operating models
Hermes and HeyBob get compared because both put an AI agent to work, but they sit on opposite sides of one question: who runs the thing? Hermes is Nous Research's open-source hermes-agent, released under the MIT license. You clone it, host it yourself, and operate it, from a $5 VPS all the way up to a GPU cluster. It works with any model, keeps persistent memory you can search with full-text queries, ships a built-in cron scheduler, can spawn subagents, and describes itself as self-improving. You talk to it through a CLI or through messaging gateways like Slack, Discord, Telegram, WhatsApp, and Signal. It is a capable base for teams that want to own every layer of the runtime.
HeyBob is the opposite operating model. It is a managed AI employee product: we run the runtime, so nobody on your team is patching a VPS or babysitting a scheduler. Someone @mentions Bob in Slack or Microsoft Teams, Bob does the work in his own sandboxed computer, pauses for an Approve/Deny button before sensitive writes, and posts an itemized receipt at the end. Personal apps default to per-user OAuth, so each query runs under the asker's own account and role rather than a shared service account.
Side by side
| HeyBob (Bob) | Hermes (hermes-agent) | |
|---|---|---|
| Who operates it | We do. Fully managed runtime; you operate the product. | You do. Self-hosted on your own infrastructure, from a $5 VPS to a GPU cluster. |
| License | Commercial product, hosted or self-hosted in your VPC on the Enterprise tier. | Open-source, MIT licensed. |
| Where work happens | Slack and Microsoft Teams, where your team already works. | A CLI plus messaging gateways: Slack, Discord, Telegram, WhatsApp, Signal. |
| Before-write behavior | External writes pause for an Approve/Deny button by default, per tool and per action. | Whatever you build into it. |
| Accounting | Every run ends with an itemized receipt (tokens, tools, sandbox time), plus a hard spend cap. | What you instrument yourself. |
| Per-user connections | Personal apps (Gmail, Zoom, HubSpot) default to per-user OAuth, so each query runs under the asker's role. | Whatever OAuth you wire up. |
| Memory | Workspace memory and per-user memory, both inspectable and editable. | Persistent memory with full-text search. |
| Model | Internal tiered model routing; you can also mandate a model per workspace. | Works with any model you point it at. |
| Pricing | Plans scoped by monthly credit volume, unused credits roll over, hard spend cap. | Your infra bill plus your time. |
Pick Hermes if you want to own the runtime end to end, your team is comfortable hosting and operating agent infrastructure, and you value an open-source, MIT-licensed base you can read and change. It runs anywhere from a cheap VPS to a GPU cluster, works with any model, and brings its own memory, scheduler, and subagents out of the box.
Pick HeyBob if you want that kind of capable agent without operating it yourself: a coworker who already lives in Slack and Microsoft Teams, pauses before writing to your systems, shows you a receipt for every run, and gives each person their own per-user connections. We run the runtime so you don't have to.
See a receipt for yourself.
Put Bob to work today, then book a call with the founder, and get 6,000 credits free.
Talk to the founder See pricing →