Skip to content

Configure Hermes

Pick an API mode, add models, and maintain Hermes model and memory configuration.

Updated View as Markdown
For humans

Hermes keeps several providers at once and stores them in config.yaml under its default directory ~/.hermes. OcHub can also manage its MCP servers and skills. Hermes may be disabled by default; enable it under Settings → Apps and plugins before opening its page.

Configure a custom connection

  1. Choose an API mode:

    Mode Upstream
    chat_completions OpenAI-compatible Chat
    anthropic_messages Claude Messages
    codex_responses OpenAI Responses
    bedrock_converse AWS Bedrock Converse
  2. Enter Base URL and API key.

  3. Add models. The first model ID also becomes the runtime and /model default.

  4. Add optional display names and context lengths.

  5. Use rate_limit_delay for a required delay between calls.

  6. Review the generated config.yaml, save, and restart Hermes.

Without an explicit mode, Hermes treats a provider as Chat Completions, so Anthropic, Responses, and Bedrock upstreams must select the matching mode.

Model and memory maintenance

Under Advanced tools → Hermes model and memory, you can:

  • Read and save the model node in config.yaml.
  • Edit MEMORY and USER memory content.
  • Toggle each memory type.
  • Inspect character counts and enabled state.

Do not edit while a running Hermes agent is writing memory. Record any backup path shown after a save or toggle.

After saving, follow the verification checklist to confirm the new connection works.

Navigation

Type to search…

↑↓ navigate↵ selectEsc close