My Hermes vs Copilot verdict: GitHub Copilot is the stronger tool for any job that ends in a pull request, Hermes Agent is the stronger tool for any job that ends in a message or a schedule, and they score high on different rows.
I rated both on eight criteria using only what each vendor's documentation says.
The numbers are editorial judgement from those docs, not a lab test.
GitHub documents Copilot at docs.github.com, and Nous Research documents Hermes at hermes-agent.nousresearch.com.
Neither vendor compares its product with the other.
For a verdict on your own stack, get your Backlink Gameplan with my team, or try Polar Rank free for 3 days.
Hermes vs Copilot scorecard
| Criterion | Hermes | Copilot | Why |
|---|---|---|---|
| Coding in the editor | 6/10 | 10/10 | Copilot lives in VS Code, Visual Studio, JetBrains, Xcode and Eclipse; Hermes reaches editors over ACP |
| Issue to pull request | 5/10 | 10/10 | Assign an issue and Copilot returns a PR; Hermes documents a webhook recipe that comments on PRs |
| Reach outside the code | 10/10 | 6/10 | Hermes answers on around 20 messaging platforms; Copilot's Slack, Teams and Mobile entry points all lead back to repos |
| Scheduled work | 9/10 | 7/10 | Hermes cron delivers anywhere; Copilot Automations need specific plans and a private or internal repo |
| Model choice | 10/10 | 7/10 | Hermes takes almost any provider or a local model; Copilot offers a model menu chosen by GitHub |
| Where it runs | 10/10 | 7/10 | Hermes runs on your hardware or six terminal backends; Copilot runs locally, in GitHub's sandbox or on Actions |
| Safety controls | 8/10 | 8/10 | Hermes gates dangerous commands; Copilot has a firewall and sandboxes, but MCP tools run without asking |
| Team administration | 5/10 | 9/10 | Copilot Business and Enterprise add central policy control; Hermes relies on allowlists you configure |
Add up the columns and the totals land close together.
That's the point: each tool wins the rows it was built for.
How I scored each row
Every row asks the same thing: what does the vendor's own documentation say the tool does out of the box?
A 10 means the feature is built in, documented and central to the product.
A middle score means you can get there, but you assemble it from parts.
A low score means the docs describe no real feature for that job.
What you're comparing
Copilot is GitHub's AI assistant for software work: completions, chat about a codebase, code review and agents that take tasks (overview).
Its agent surfaces are Copilot CLI, the desktop Copilot app, and the Copilot cloud agent (formerly the Copilot coding agent) on github.com.
Hermes is an MIT-licensed general agent on GitHub with a CLI, a TUI, a Desktop app, a messaging gateway, cron and a learning loop that writes skills from finished work.
One is shaped around a repository.
The other is shaped around a person and their week.
Where Copilot is strongest
- The full GitHub loop. An issue can be assigned to Copilot, which raises a pull request and requests your review when it's done.
- Merge blockers. Agent Merge's Watch and fix, added to pull requests on 28 October 2026, has Copilot fix failing Actions workflows or merge conflicts (changelog).
- Many entry points. The same cloud session can be started from the web, GitHub Mobile, the CLI, the Copilot app, VS Code, Slack or Teams.
- A careful CLI. Plan mode asks clarifying questions and builds a plan before it writes code, and
/sandbox enablefences off your filesystem and network. - Shared customization. Custom agents live in
.github/agents, and skills follow the open Agent Skills standard.
Where Copilot falls short
- It only works with repositories hosted on GitHub, and it isn't offered for GitHub Enterprise Server.
- Once a repo admin configures an MCP server, the cloud agent uses its tools without asking for approval (MCP docs).
- Copilot connectors don't yet work in Actions sessions, which includes the classic assign-an-issue flow.
- The cloud agent can't see comments added to an issue after assignment.
- Issue assignment and the agent tasks API are both still in public preview.
- Outside software, it has nothing to offer: no reminders, no inbox, no general errands.
🔥 Want the Hermes side set up properly, not guessed at? Inside the AI Profit Boardroom you get step-by-step video trainings, the installable Agent OS, four coaching calls a week and 3,400+ members testing agent setups side by side. → Join the AI Profit Boardroom
Where Hermes is strongest
- One gateway, many apps. Telegram, Discord, Slack, WhatsApp, Signal, SMS, email, Microsoft Teams and more connect through a single background process (messaging docs).
- Schedules in plain English. Cron jobs can attach skills, pin a model, deliver to any connected platform, or fire from a webhook.
- Any model.
hermes modelswitches providers with no code changes, including your own endpoint. - Memory and skills. It keeps notes about you across sessions and saves repeat procedures as skills in the agentskills.io format.
- One history. Since v0.23.0, a single gateway owns every session, so a chat started on your phone is the same chat in the terminal (release notes).
Where Hermes falls short
- It has no native issue-to-pull-request agent; the closest documented recipe is a webhook that reviews a diff and posts a comment.
- Its approval system covers dangerous commands, not outgoing messages, so customer-facing jobs need "draft only" in their instructions.
- You host it, which means you also keep it running, updated and backed up.
- Inside an editor, ACP mode drops messaging delivery and cron management, so it's a narrower Hermes than the gateway one.
- There's no central admin console comparable to Copilot Business policies in the docs I read.
Instructions, skills and memory compared
| Feature | Hermes | Copilot |
|---|---|---|
| Project rules | Loads AGENTS.md from the working folder; /init can write one from a repo scan | .github/copilot-instructions.md, path-specific .instructions.md files, and AGENTS.md, CLAUDE.md or GEMINI.md |
| Skills | Stored in ~/.hermes/skills/, agentskills.io format, and the agent writes new ones | Stored in .github/skills, .claude/skills or .agents/skills |
| Specialist agents | Profiles and Bot Mode, each with its own role and model | Custom agent profiles in .github/agents |
| Memory | Curated memory files plus search over past sessions | Copilot Memory: facts about a repository and your personal coding preferences, on paid plans |
Both read AGENTS.md, which is the simplest way to make them agree on house rules.
GitHub adds a caveat: not every Copilot feature reads agent instruction files yet, so keep the critical rules in copilot-instructions.md as well.
On skills, the two formats are close enough that a well-written procedure can live in both places with light edits.
Setup effort, compared honestly
Copilot asks little of an owner: a plan, a GitHub repo, and a developer who turns on the parts they need.
The cloud sandbox is hosted by GitHub, and the CLI installs with one npm, WinGet or Homebrew command.
Hermes asks more.
You install it, choose a model provider with hermes model, connect chat apps with hermes gateway setup, and keep that machine running.
The reward for that effort is control: the data, the model, the schedule and the approval rules all sit with you.
Weigh that trade before you weigh any feature list.
The one documented link between them
Hermes can run on your Copilot subscription.
Version 0.23.0 moved Copilot support out of Hermes core into Nous Research's github-copilot labs plugin, installed with hermes plugins install github-copilot (providers page).
It adds a copilot provider for the Copilot API and a copilot-acp provider that drives your installed Copilot CLI over ACP.
The plugin README says Nous built it and doesn't support it, and the catalog lists it in the community tier.
So it earns a mention in the verdict, but not extra points.
Three jobs, three winners
An online shop owner who wants a morning briefing
Hermes wins outright.
A cron job researches overnight supplier news and sends a short summary to the owner's Telegram at 7am; Copilot has no equivalent because there's no repo.
A design studio fixing a client's site
Copilot wins.
The studio assigns the bug to Copilot on GitHub, reviews the pull request, and switches on Watch and fix so failing checks get handled before merge.
A bookkeeping firm that wants both
It's a split.
The partner who maintains the firm's export script uses Copilot CLI in plan mode, while Hermes posts monthly deadline reminders in Microsoft Teams and drafts client emails that staff send themselves.
If your shortlist also includes a code editor, see my Hermes vs Cursor verdict.
📞 Want a verdict on your own tool stack? Get your Backlink Gameplan and my team will review where each agent belongs in your business, or try Polar Rank free for 3 days. Backlink Gameplan or try Polar Rank free for 3 days
Who should pick Copilot, and who shouldn't
Pick it if your code is on GitHub, your developers live in a supported editor, and you want pull requests, reviews and repo automations with admin controls.
Skip it if you have no codebase, your code sits on another host, or you need GitHub Enterprise Server.
Who should pick Hermes, and who shouldn't
Pick it if you want one assistant across your chat apps, scheduled reports, memory of how you work, and full control over the model and the machine.
Skip it if nobody on your team will look after a server, or if all you want is help writing code inside an editor.
Final verdict
Copilot is the better coding product, and it isn't close for repo work.
Hermes is the better general agent, and it isn't close for chats and schedules.
If you can take on only one setup project this quarter, start with the lane where most of your hours go: code or communication.
A team that spends its day in pull requests gets more from Copilot first.
A team that spends its day answering people and chasing deadlines gets more from Hermes first.
For a business that ships software, the strongest setup in the docs is both: Copilot on the repository, Hermes on the gateway, with the labs plugin connecting them where that helps.
🔥 Get the playbooks for running both Inside the AI Profit Boardroom you get step-by-step video trainings, the installable Agent OS, four coaching calls a week and 3,400+ members sharing what works. → Join the AI Profit Boardroom
FAQ: Hermes vs Copilot Verdict
Which scores higher overall, Hermes or Copilot?
The totals land close because each wins different rows. Copilot leads on editor and pull request work, Hermes on reach, scheduling and model choice.
Did you benchmark the two tools?
No. The ratings are editorial judgement based on what GitHub and Nous Research document for each product.
Is the Hermes github-copilot plugin supported?
Nous Research built it but labels it unsupported, and the plugin catalog lists it in the community tier.
Can Copilot work on code hosted outside GitHub?
No. GitHub says Copilot only works with repositories hosted on GitHub.
Is Hermes a replacement for Copilot?
Not for repo work. Hermes has no native issue-to-pull-request agent, so it fits better as the assistant around your code.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I've built a 7-figure SEO and link building agency, Goldie Agency, with a team of 70+ SEO experts.
- I've grown a YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught over 29,000 students.
→ Get my best AI training inside the AI Profit Boardroom
Score your own jobs against the table, then get your Backlink Gameplan if you want my team to sanity-check your hermes vs copilot decision, or try Polar Rank free for 3 days.











