Hermes Agent vs Orca: Completion Gates or Pull Request Repair?

6 min read Comparisons

Compare Hermes Agent and Orca on goal checks, pull request feedback, recurring triggers and who operates the agent computer.

Hermes Agent and Orca serve developers who need agents to finish work, but their published delivery loops point to different decisions. Hermes Agent documents completion contracts and independent review skills, while Orca documents sending pull request comments and failed checks back to an agent for repair.

Choose around the point where your work usually stalls. A job with an unclear finish line needs explicit completion evidence. A branch waiting on review comments needs a route from those comments back to the agent responsible for the change.

Hermes Agent defines a finish line

Hermes Agent's goal guide describes persistent objectives with a judge, completion contracts and shell quality gates. A test command can become a condition that must pass before the agent considers a goal complete. That gives a developer a concrete place to define success before the work begins.

Its Kanban guide adds a root card owned by an orchestrator profile, child cards for specialists and completion judgment that can create more work. Comments, dependencies and attempts stay with the cards. People can manage and discuss that work, but the published card assignee is a profile name.

The same board has boundaries worth checking before adopting it for a team. Its documented trust model is local and single-host. Dependency handoffs work, while workflow-template routing fields are reserved for a future version and are ignored by the kernel.

Orca works through the coding workspace

Orca brings agent chats, terminals, worktrees and review into a coding environment. Its GitHub review documentation describes exposing failed hosted checks and sending their names and links to an agent. Pull request review comments can also go back for correction.

That is useful when a maintenance change already has a branch and reviewers have identified the next action. Repository hooks run during commit, with failures shown inline in its commit and push guide. Those checks provide evidence from the repository's existing tooling.

Orca also publishes coordinator-and-worker orchestration with Tasks, dependencies and blocking questions. Its orchestration documentation labels this experimental and says the orchestration panel and switch are temporarily unavailable. Keep that documented command-line process separate from an available desktop workflow builder when evaluating it.

Compare review checks at the same stage

Hermes Agent publishes a pre-commit review skill that delegates security and logic review to a separate, diff-only reviewer in interactive sessions. Its independent Kanban review checks handoffs against evidence and keeps the reviewer from editing the implementation. These are published mechanisms for a second perspective on completed work.

Orca's diff annotation guide describes batching human comments and sending them for agent revision. Its public pages do not describe an independent second review specifically for risky changes. A human diff review, a repository hook and an independent agent review answer different questions, even if each can find a defect.

The delivery evidence also differs after a pull request opens. Orca documents returning feedback and failing checks to the agent. Hermes Agent documents opening pull requests and reading comments, but its cited public pages do not establish a review-feedback revision cycle. If review churn is your recurring bottleneck, make that cycle part of the evaluation instead of inferring it from the ability to open a pull request.

Separate scheduled work from incoming events

Hermes Agent's cron guide describes scheduled jobs carrying their own instructions, skills and delivery settings. Its webhook routes can trigger those same jobs from outside events. On a self-hosted installation, agent-backed jobs need a running gateway, and a scheduler tick skips them when it is down.

Orca's automation guide describes hourly, daily, weekly, cron and RRULE schedules, with timezones and run history. Its public pages do not describe event or webhook triggers for those automations. It can separately launch coding work from an external issue through a workspace.

For a recurring test-maintenance job, either schedule may be relevant. For a job that should begin immediately when a service sends an event, Hermes Agent has a documented webhook route. Do not treat the existence of an issue integration as proof of a general automation event trigger.

Locate the computer and the bill

Hermes Agent's core software is free under MIT, with optional hosted execution. Hermes Cloud lists Plus $20/month, Super $100/month and Ultra $200/month, plus credit top-ups. Its FAQ separately meters compute and storage, excluding inference and tools, and does not explain how that relates to included plan hosting.

Orca's source repository describes free MIT-licensed software. Its execution guide offers local machines, SSH hosts and self-hosted servers, with experimental Cloud VM recipes using your provider account. Agent subscriptions and provider charges are separate, and Orca says it does not sell managed VPS hosting.

Both can therefore fit a developer operating their own agent computer. Optional Hermes Cloud and Orca's customer-controlled remote machines carry different operational responsibilities. Decide whether you intend to maintain that machine before comparing the software prices.

Hermes Agent and Orca, point by point

This table compares the same questions for both products, with a source and check date for each answer. It preserves experimental status and the boundaries of what their public pages describe.

Both columns answer the same questions from each company's public pages. Yes means the company publishes the capability the question asks about, Partly means something narrower, No means it says it does not offer it, and Not published means its public pages do not describe it. Every answer links to its source and shows when we last checked it. The oldest check on this page was November 13, 2026.

Hermes Agent and Orca compared, with the source and check date of each answer
What you compare Hermes Agent Orca
Goals Hermes Agent Partly

Persistent goals keep an agent working toward an objective, with a judge and optional completion tests. Its docs describe no linked metric tracking against an ongoing business goal.

Orca Partly

Codex chats display an active goal, its state and controls to pause or resume it. Its docs do not describe an owner tracking a measurable target against that goal.

orca.dev/…/native-chat
Leads Hermes Agent Yes

You can name an orchestrator profile to own a decomposed project's root task, assign child tasks to specialists, judge completion and add more work when needed.

Orca Partly

Experimental orchestration gives a coordinator supervised workers, completion messages and blocking questions. Its docs describe coordination within a Run rather than a standing project, goal or team lead role.

orca.dev/…/orchestration
Tasks Hermes Agent Partly

Kanban cards keep an assigned agent profile, comments, dependencies and attempt history. People can comment and manage cards, but the documented assignee is a profile name rather than a person or an agent.

Orca Partly

Orca displays external issues and experimental orchestration Tasks with specs, dependencies, status and worker attempts. Its docs do not establish the same work record assignable to either a person or an agent.

Recurring work Hermes Agent Yes

Cron jobs run once or recur on a schedule, and webhook events can fire the same jobs immediately. Each job can carry its own instructions, skills and delivery settings.

Orca Partly

Automations run prompts on hourly, daily, weekly, cron or RRULE schedules, with timezones and run history. Its docs do not describe incoming event or webhook triggers for those automations.

orca.dev/…/automations
Workflows Hermes Agent Partly

Kanban dependency chains coordinate roles such as scout, editor and writer. They provide ordered handoffs, but the kernel does not execute workflow templates, whose routing fields are reserved for a future version.

Orca Partly

Experimental orchestration supports task dependencies and decision gates that block tasks. Its docs do not describe a saved multi-step process that repeats unchanged, and the orchestration panel and switch are temporarily unavailable.

orca.dev/…/orchestration
Autonomy Hermes Agent Yes

You configure shell approval modes, permanent allowlists and deny rules. Dangerous-command requests appear in the terminal or messaging chat, where you can approve or deny them.

Orca Yes

Choose Manual, Auto or Yolo permissions for new agent sessions. Structured chat presents approval cards, alerts you when input is needed and retains answered cards as receipts.

Learned autonomy Hermes Agent Yes

The approvals suggest command mines executed, approved commands into ranked allowlist proposals. You choose which entries to apply, and destructive command classes are excluded.

Orca Not published

Its docs describe permission presets and approval receipts, but do not describe Orca proposing permission changes from your approval history.

Memory Hermes Agent Yes

Persistent notes and user preferences carry into later sessions. You can inspect, edit or delete learned memory through the journey commands or the desktop panel.

Orca Partly

Orca respects agent-owned CLAUDE.md and AGENTS.md files and lets you edit them inline. Its docs do not describe Orca saving lessons from corrections for later runs.

orca.dev/…/hooks-memory
Planning Hermes Agent Yes

From chat, Kanban can expand a rough idea into a specification or dependency-linked tasks. You can disable automatic child promotion so the generated tasks wait for manual review before execution.

Orca Yes

Claude and Codex structured chats have Plan mode. Their plan updates render as inline checklists, with the latest plan in a progress panel above the composer.

orca.dev/…/native-chat
Intake Hermes Agent Yes

Signed webhooks turn outside events into agent prompts. A route can target a named profile when profile multiplexing is enabled, and filter which events start work.

Orca Yes

You can launch an agent from a Linear issue, passing its description, comments, images and sub-issues into the prompt. This routes an external issue through a workspace rather than a general email intake address.

orca.dev/…/linear
Connectors Hermes Agent Yes

A curated Model Context Protocol catalog and custom local or remote servers connect external tools. HTTP webhook routes accept signed service events and can post responses back through integrations such as GitHub.

Orca Partly

Orca connects issue providers and lets you register Model Context Protocol servers for compatible agents. Its docs do not describe event or webhook triggers alongside those connections.

Models Hermes Agent Partly

You choose the main model and can override auxiliary tasks. Provider routing chooses a backend for a selected model, rather than selecting a model from each task's needs.

Orca Partly

You choose models and effort in chat or set model overrides for individual workers. Its docs do not describe Orca automatically selecting a model for each task or step.

Cloud or your own computer Hermes Agent Yes

Run the free agent locally through its desktop or terminal interfaces, or choose the optional Hermes Cloud service for a hosted instance with a persistent workspace.

Orca Partly

Run locally, on SSH hosts or on a self-hosted Orca server. Experimental Cloud VM recipes use your own provider account. Orca says it does not sell managed VPS hosting, so these are customer-controlled compute options.

orca.dev/docs/ways-to-run
Your own subscription Hermes Agent Yes

Hermes accepts your provider API keys and a ChatGPT account login. Its Claude account path requires Max plus purchased extra-usage credits, not the included Max allowance.

Orca Yes

Orca runs existing agent subscriptions side by side, including Claude Code and Codex. You bring the subscriptions rather than buying model usage from Orca.

orca.dev
Pricing Hermes Agent Yes

The MIT agent is free. Portal lists Plus $20/month, Super $100/month and Ultra $200/month, with credit top-ups. Cloud's FAQ separately meters compute and storage, excluding inference and tools, without explaining how that relates to included plan hosting.

Orca Partly

Orca is free and open source under the MIT License, rather than a usage or credit plan. Agent subscriptions and any cloud provider charges are separate costs you supply.

Correctness Hermes Agent Yes

Goal completion contracts define verification criteria, and shell quality gates must pass before a goal completes. Kanban review also maps acceptance criteria to evidence and can return work for correction.

Orca Yes

Repository pre-commit hooks run normally, with failures shown inline. Hosted checks expose failed jobs, and Fix broken checks passes their names and links to an agent for repair.

For software work

Hermes Agent and Orca compared, with the source and check date of each answer
What you compare Hermes Agent Orca
Code review rules Hermes Agent Yes

The pull-request review guide uses reusable review guidelines and remembered team conventions, such as required type annotations and database access rules, to check later diffs.

Orca Partly

Orca documents a human diff review for necessity, minimal changes and consistency, then batches comments for agent revision. Its docs do not establish automatic code review against configured repository rules.

Adversarial review Hermes Agent Yes

A bundled pre-commit skill delegates security and logic review to a separate, diff-only reviewer in interactive sessions. Kanban's review skill independently verifies handoffs and keeps the reviewer from editing the implementation.

Orca Not published

Its docs describe parallel agents and human diff review, but do not describe an independent second review of risky changes.

Risk rating Hermes Agent Not published

Its docs describe no change-level risk rating before merge. Optional residual-risk notes on Kanban handoffs and severity labels for individual review findings do not establish such a rating.

Orca Not published

Its docs show hosted check results, review state and merge controls, but do not describe a change risk rating before merge.

Pull requests Hermes Agent Partly

Agents can open pull requests through the GitHub CLI, and the bundled GitHub skill requires reading PR comments. The cited docs do not establish a review-feedback revision cycle.

Orca Yes

Orca drafts pull request details from the diff and sends failing checks or pull request review comments back to an agent to fix. Hosted reviews and comment threads are accessible inside the app.

A fact we cannot find or cannot check is left out, and prices can change after the date shown. Task Machine publishes this comparison and is an interested party. Hermes Agent and Orca are named only to identify them, and there is no relationship between the companies. Has something changed? Tell us.

Who each one suits

Pick Hermes Agent if your priority is defining a goal contract, coordinating profiles and reviewing completion evidence. Its limit is that opening and reading pull requests does not establish the feedback-revision cycle described by Orca.

Pick Orca if you want parallel coding workspaces and a documented route from pull request comments or failed checks to agent fixes. Its orchestration remains experimental, with the panel and switch temporarily unavailable, and you supply remote compute.

Where Task Machine fits

Task Machine connects the finish line and delivery record through a Task's acceptance requirements. Its review documentation describes checks against delivered work before human review, and its pull request process records approval against the exact commit. A person or agent can own the Task containing the plan, decisions and runs.

Task Machine is the wrong pick if free software for a hands-on coding workspace is your entire requirement. Use Task Machine vs Hermes Agent and Task Machine vs Orca to compare ownership and review boundaries, then select one stalled pull request and trace who receives its next failed check.