Hermes Agentvs OpenCode

OpenCode is an open source coding agent from Anomaly. You work with it in a project through the terminal, the desktop app, a password-protected web UI or editors such as Zed over ACP, and an iPhone app in public beta steers sessions running on your computer. OpenCode Console, its team service, adds hosted agents that its privacy policy lists for Slack, GitHub, Discord and email; the Slack agent, which answers mentions in a thread, is in early access.

Hermes Agent answers on more than 20 messaging platforms, including Telegram, Discord, Slack, WhatsApp, Signal, Microsoft Teams and iMessage through BlueBubbles. One gateway process you run connects them all, and by default it accepts only users you allowlist or pair. Hermes Agent can send spoken replies in Telegram and Discord and talk in Discord voice channels.

Outside chat apps, you can use Hermes Agent in Hermes Desktop on macOS (Apple Silicon), Windows and Linux, in the Webapp in any browser, in the terminal, and in ACP-compatible editors such as VS Code, Zed and JetBrains.

The Hermes Agent gateway runs a built-in scheduler. You write a schedule in plain language or cron syntax, each job runs in its own agent session, and results go back to the chat where you created the job, to local files or to other platforms.

Jobs run while the gateway runs, so on a server you keep on or on Hermes Cloud they keep going with your laptop closed. Hermes Cloud runs Hermes Agent on a dedicated, always-on instance with its own workspace and dashboard, in the nearest available region, and every paid Nous Portal plan includes one Hermes Cloud agent.

OpenCode added scheduled prompts in October 2026 as an experimental feature you turn on with experimental.schedules. Enabled schedules run under opencode serve, and scripts and CI jobs can start work with opencode run.

Hermes Agent keeps a small, curated memory of your preferences, projects, environment and what it has learned, and loads it at the start of every session. It stores every CLI and chat session in SQLite with full-text search.

When it works out a non-trivial workflow, it saves the approach as a skill and improves that skill as it uses it. A write-approval gate can hold skill changes for your review, and a background curator maintains the skills it creates.

OpenCode gets persistent project guidance, such as build commands and code conventions, from AGENTS.md files you add, and you write its skills yourself as SKILL.md files. Hermes Agent reads AGENTS.md project files as well.

One Nous Portal login gives Hermes Agent hundreds of models, with new frontier models the day they launch. Paid Nous Portal plans include Hermes Agent usage within limits that reset daily, weekly and monthly, and hosted tools: web search, image generation, speech and a cloud browser.

Hermes Agent also works with any OpenAI-compatible API, OpenRouter, ChatGPT sign-in for Codex models, SuperGrok sign-in, and local models through Ollama, vLLM, llama.cpp or SGLang. You switch models mid-session with /model, and conversation history, memory and skills carry over. Fallback providers are opt-in.

OpenCode works with 200+ providers through models.dev and with local models, and it signs in with ChatGPT Plus or Pro and GitHub Copilot.

An official optional Hermes Agent skill hands coding work to the OpenCode CLI, for features and PR reviews; install it with hermes skills install official/autonomous-ai-agents/opencode. Hermes Agent also includes OpenCode Zen (pay-as-you-go) and OpenCode Go as model providers, so it can use the models you pay for through OpenCode.

OpenCode's docs list Hermes Agent among the harnesses with both built in. They say Hermes Agent picks the right API for each paid or Go model, and that versions v0.21.6 and later send the session ID that OpenCode Go requires. OpenCode's free models work only inside OpenCode, so Hermes Agent cannot use them.

Teams on OpenCode Console can also connect Hermes Agent with npx @opencode/console setup hermes. Console's guide says your Hermes Agent bots and cron jobs keep working, under the workspace's models and policies.