Verify Kimi K2.5 in OpenClaw
Use Kimi K2.5 in OpenClaw only if your Moonshot account can still complete a live request with that exact model. If your workload accepts a replacement, fresh Moonshot onboarding selects Kimi K3.
Find out whether the exact model is required
Treat Kimi K2.5 as an exact requirement if you are reproducing a prior result or maintaining a workflow tied to that model. Check your Moonshot account's model offerings for the intended endpoint region and verify live access under that same account. The supplied OpenClaw provider documentation establishes Kimi K3 as the fresh onboarding default. Confirm any K2.5 availability or retirement date with Moonshot before scheduling work that depends on it.
If K2.5 is a preference, start with a disposable test instance or session for Kimi K3. If exact K2.5 access is essential and Moonshot cannot confirm it, keep the existing workload in place and record the model dependency before changing the default.
Give each endpoint and model combination its own trial alias. A credential accepted on an endpoint does not establish access to every model name.
Moonshot AI - OpenClawProve the model on a real request
Inspect the model reported by /status in the session you will use for testing. Then send a short task with a known result and check /status again after the reply. Capture the displayed model and whether the expected result came back. Use the reported model and any available provider-side request records to investigate a reply attributed to a different route.
For an illustrative migration, suppose your test input is a 12-ticket fixture and your expected output is a JSON array containing exactly the three IDs marked urgent. Before promotion, require the correct IDs and a session model of moonshot/kimi-k3. If you are evaluating exact K2.5 compatibility, use the same fixture and additionally require evidence of a live response from that exact model.
Record the model reported on the response before accepting the route. Keep archived tests for an old model separate from trials of its successor.
API usage and costs - OpenClawMatch the credential to the endpoint region
For a fresh Moonshot setup, run openclaw plugins install @openclaw/moonshot-provider on the instance where OpenClaw runs. Use openclaw onboard --auth-choice moonshot-api-key for the international endpoint or openclaw onboard --auth-choice moonshot-api-key-cn for the China endpoint. Choose the endpoint that matches the Moonshot account you intend to use and provide its credential through the authentication flow.
Run openclaw models list --provider moonshot after setup and record the exact model reference you intend to test. Kimi Coding uses a separate provider route, documented as kimi/kimi-for-coding. If a credential is rejected, verify which service issued it and which endpoint the setup selected before changing model names.
Moonshot AI - OpenClawPromote Kimi K3 after the pilot passes
For an existing Moonshot installation, record the previous default and where your deployment manages its model configuration before making a change. OpenClaw documents openclaw models set moonshot/kimi-k3 for selecting Kimi K3 explicitly. After setting it, open a fresh session and repeat the same fixture. A successful cutover shows the intended model in the new session and still meets the fixture's acceptance criteria.
An upgrade can require workflow changes even when authentication succeeds. For example, if the reply has the correct IDs but wraps the JSON in extra text, the parser may still reject it. Restore the previous working default if it remains available, or leave the affected workflow on hold while you check the output contract in the pilot.
Moonshot AI - OpenClaw