⌁ Guides

OpenHarness

Register Gondola as a compatible-endpoint provider in HKUDS's OpenHarness with one oh provider add command, and run any Venice model paid per request in USDC on Base.

Create an API key

Sign in with your Base wallet at /keys and mint a Gondola API key. Keys start with gnd_. It authenticates every request and each call is charged in USDC against your Gondola balance.

Top up USDC on Base once at /wallet. The same balance powers the browser chat and every tool you connect.

Add the provider

OpenHarness (the oh CLI, with the built-in ohmo personal agent) treats providers as named profiles. Register Gondola as an OpenAI-compatible profile:

macOS / Linux / WSL

oh provider add gondola \
  --label "Gondola" \
  --provider openai \
  --api-format openai \
  --auth-source openai_api_key \
  --model claude-sonnet-5 \
  --base-url https://api.gondola-ai.com/v1

Windows (PowerShell)

openh provider add gondola `
  --label "Gondola" `
  --provider openai `
  --api-format openai `
  --auth-source openai_api_key `
  --model claude-sonnet-5 `
  --base-url https://api.gondola-ai.com/v1

Windows: oh is a PowerShell alias collision

In PowerShell, oh resolves to the built-in Out-Host alias, not the CLI. Use openh on Windows instead; macOS, Linux, and WSL keep using oh.

Switch to the new profile and confirm it authenticated:

shell

oh provider use gondola
oh provider list

Pasting the key

oh provider add only registers the profile; it has no key yet. Switching to it (or picking it from oh setup) prompts you to paste the Gondola key. oh provider list shows [missing auth] next to a profile until that happens, flipping to [ready] once it has a key. Credentials are stored per profile, so pasting the Gondola key here never touches a key you already have saved for a real OpenAI profile.

Pick a model

The command above defaults the profile to claude-sonnet-5, a strong general-purpose pick. For a coding specialist, use qwen3-coder-480b-a35b-instruct-turbo instead, either by re-running oh provider add with a different --model, or per run with the global -m/--model flag on oh. Browse the full catalog, with prices and context windows, any time:

shell

curl https://api.gondola-ai.com/v1/models

Prompt privacy

Gondola never stores prompts or completions: requests are metered, forwarded, and gone when the response finishes (the one exception, a video source clip over 18 MiB, is held privately for the length of its job; see /privacy), and Venice runs a no-logging model underneath. For zero retention on Venice-controlled infrastructure too, pick a model whose privacy field reads private on GET /v1/models. Details on /privacy.

Notes and troubleshooting

oh --dry-run, not oh doctor

OpenHarness has no separate doctor command. Use oh --dry-run to preview a session (auth state, resolved settings, MCP config) without calling the model; it reports a ready/warning/blocked verdict with a concrete next step.

Capping spend on unattended runs

OpenHarness's compatible-endpoint profiles don't document a way to set a custom request header, so x-gondola-max-factor and x-gondola-max-charge-usd aren't reachable from oh provider add. For unattended ohmo sessions, cap spend at the key instead: mint a dedicated key with a spend limit at /dashboard?tab=keys, editable any time.

If a request returns insufficient balance, top up at /wallet. The model catalog above is live, so a model added to Venice today is callable from Gondola the same day.

Loading page

openh provider add gondola ` --label "Gondola" ` --provider openai ` --api-format openai ` --auth-source openai_api_key ` --model claude-sonnet-5 ` --base-url https://api.gondola-ai.com/v1