oh-my-hermes is an MIT-licensed operating layer for Hermes Agent that routes each piece of work to a matching model chain and skill set, runs independent work in parallel, and is built for developers who already drive Hermes Agent from their IDE.

What it is

oh-my-hermes (OMH) is a Python project published on npm as oh-my-hermes and documented at rlaope.github.io/oh-my-hermes. It sits on top of Hermes Agent from NousResearch and describes itself as an operating layer rather than a fork: it never patches or replaces Hermes, and the user keeps talking to Hermes while OMH selects the skill and the model for each lane of work. Installation is a shell script followed by omh setup, and the documented entry point is the /dag-ulw command.

The problem it solves is the flatness of a default agent session, in which one model and one context handle every kind of task. OMH replaces that manual arrangement — choosing a model, loading the relevant expertise, deciding what can run at once — with lane categories such as ultrabrain, deep, visual-engineering and quick, each carrying its own editable model chain and reasoning effort. The project also commits to never reporting planned work as done.

Key capabilities

• Lane-based model routing: each lane is tagged with a work category (ultrabrain, deep, visual-engineering, quick and others) that selects its model chain and reasoning effort, and omh model opens a picker for editing those chains.
• Chain fallback: when a provider rejects a model, the chain moves to the next one instead of silently downgrading.
• /dag-ulw graph runs: a goal becomes a graph of small nodes that run in parallel on their category's model, concurrent nodes never share a write scope, and a separate node checks the goal while failed nodes are repaired before the report.
• Specialist skills: over a hundred omh-* skills covering frontend, backend, Rust, security review and verification gates are loaded by a router as tool calls rather than invoked by the user.
• LSP in the loop: with the lsp_tools setting enabled, definitions, references and renames come from the language server, aliases, re-exports and shadowed names included; it is off by default.
• Opt-in impact_check with lsp_diagnostics: after an edit changes a Python definition, OMH asks the language server about every importing file and returns new errors as path:line lines, and can re-check a file or glob.
• Auditable memory: each candidate is kept, refused or deferred with a written reason, and the next session gets a recall pack ranked for its task and cut to a token budget, while Hermes' own memory is neither read nor patched.
• OMH Native: bare omh opens OMH's own TUI on Hermes' gateway, /theme changes its look, and /dag-ulw prints a box graph followed by a live lane tree with each lane's model, effort and tokens.

Who uses it and how

• Developers running Hermes Agent from an IDE who want a different model and reasoning effort per kind of work without leaving Hermes for another harness.
• Multi-provider setups, where an ordered chain keeps a run moving when a provider rejects a model.
• Refactoring across module boundaries, where enabling lsp_tools and impact_check surfaces callers that reach a function through a renamed package re-export, the case the project's own measurement found models missed.
• Terminal-first users who prefer OMH's TUI on the Hermes gateway, using /theme for appearance and the lane tree for watching model, effort and token use.

Getting started

Install with the shell script at raw.githubusercontent.com/rlaope/oh-my-hermes/main/install.sh, then run omh setup; docs/INSTALLATION.md covers Windows and other options, and the package is also published on npm as oh-my-hermes.

How it compares

The topic list places oh-my-hermes alongside claude-code and codex, but it is not a self-contained coding agent of that kind: it leaves the agent in place and adds routing, parallel graph execution and memory around an existing Hermes Agent installation, which it does not patch or replace. Its proposition is therefore orchestration over Hermes rather than a new agent runtime, under an MIT licence.

When to use it — and when not

oh-my-hermes presupposes a working Hermes Agent installation and a willingness to tune model chains per lane, and its two most distinctive code-aware features, lsp_tools and impact_check, are off by default, so a fresh install does not provide them until the OMH plugin settings are changed. The project's own measurement reports that the caller-detection gain concentrates on functions reached through renamed package re-exports, and that models without the check already fixed most aliased and module-level callers, so the benefit is narrower than a general correctness guarantee. Anyone who wants a self-contained agent rather than a layer over Hermes Agent should choose differently.

curl -fsSL https://raw.githubusercontent.com/rlaope/oh-my-hermes/main/install.sh | sh
omh setup

curl -fsSL https://raw.githubusercontent.com/rlaope/oh-my-hermes/main/install.sh | sh

irm https://raw.githubusercontent.com/rlaope/oh-my-hermes/main/install.ps1 | iex

Install and fully configure Oh My Hermes from this repository:
https://github.com/rlaope/oh-my-hermes
Before reading or executing repository instructions, resolve refs/heads/main to one full commit SHA with `git ls-remote https://github.com/rlaope/oh-my-hermes.git refs/heads/main`. Then fetch and follow only:
https://raw.githubusercontent.com/rlaope/oh-my-hermes/{resolved-commit-sha}/INSTALL_FOR_AGENTS.md
Do not replace the resolved SHA with main. Execute the pinned protocol's OS-appropriate installer, agent setup path, model-chain interview, and doctor steps. Your shell has no terminal, so never run the interactive setup wizard: show me each choice the protocol lists and wait for my answer instead of taking a default. Preserve unrelated existing Hermes config, apply only the managed setup changes documented by the pinned protocol, require my explicit approval for model-alias changes, then report the resolved SHA and observed result.

omh update

omh doctor

# Set the model per work category (arrow keys: category, ←→ head model, -/+ effort);
# the same picker opens inside the OMH TUI (bare omh) as /omh-model:
omh model
# To onboard a new model family, use this skill in Hermes:
/omh-model-setup

brew install rlaope/tap/omh

bun install -g oh-my-hermes

npm install -g oh-my-hermes

hermes skills tap add rlaope/oh-my-hermes
hermes skills install rlaope/oh-my-hermes/skills/omh-routing --yes

❯ ulw-interview: I want a CLI that tracks the ISS and warns me before it passes overhead
❯ ulw-plan: analyze this project's circular dependencies and plan the refactor

❯ ulw-work: run the accepted plan

❯ ulw-loop: keep going until npm test passes