kimi | Friends of the Crustacean 🦞🤝 | Page 1

lofty obsidian Apr 13, 2026, 1:15 PM

#

linked existing openclaw to kimi-claw using their script but my agent doesnt talk to it

daring ravineBOT Apr 13, 2026, 1:15 PM

#

twilit mesaBOT Apr 13, 2026, 1:15 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

daring ravineBOT Apr 13, 2026, 1:15 PM

#

knotty brookBOT Apr 13, 2026, 1:16 PM

#

First thing, paste these:

openclaw --version
openclaw status
openclaw models status
openclaw config get agents.defaults.model

If the chat is totally silent, also add:

openclaw channels status --probe
openclaw logs --follow

Most common issue here: the script adds Kimi auth/provider, but your agent is still using some other default model.

#

Also double check which Kimi path you wired up, because these are separate:

Moonshot K2 → MOONSHOT_API_KEY + model like moonshot/kimi-k2.5
Kimi Coding → KIMI_API_KEY + model like kimi/kimi-code

Those keys/model refs are not interchangeable.

Docs:

FAQ: https://docs.openclaw.ai/help/faq
Troubleshooting: https://docs.openclaw.ai/gateway/troubleshooting
Moonshot/Kimi setup: https://docs.openclaw.ai/providers/moonshot

Paste the outputs and I’ll help narrow it down fast.

lofty obsidian Apr 13, 2026, 1:56 PM

#

got an api key from nvidia, how can i use it?

knotty brookBOT Apr 13, 2026, 1:57 PM

#

Yep, easiest path is:

export NVIDIA_API_KEY="nvapi-..."
pnpm openclaw onboard --auth-choice skip
pnpm openclaw models set nvidia/moonshotai/kimi-k2.5
pnpm openclaw models status

A couple important bits:

just setting the key is not enough, you also need to switch your agent’s default model
if you want Kimi through NVIDIA, use:
- nvidia/moonshotai/kimi-k2.5
if you want NVIDIA’s default hosted model instead, use:
- nvidia/nvidia/nemotron-3-super-120b-a12b

#

If you want it persisted in config, the key is:

env: { NVIDIA_API_KEY: "nvapi-..." }

Docs:

NVIDIA provider: https://docs.openclaw.ai/providers/nvidia
Models CLI: https://docs.openclaw.ai/cli/models
FAQ: https://docs.openclaw.ai/help/faq

If you run those and it still doesn’t pick up Kimi, paste:

pnpm openclaw --version
pnpm openclaw models status

#

pnpm openclaw config get agents.defaults.model

lofty obsidian Apr 13, 2026, 1:57 PM

#

im gonna be using minimax 2.7

knotty brookBOT Apr 13, 2026, 1:58 PM

#

Then I’d skip the NVIDIA key for this part.

If you want MiniMax M2.7, use the MiniMax provider directly:

export MINIMAX_API_KEY="sk-..."
pnpm openclaw onboard --auth-choice minimax-global-api
pnpm openclaw models set minimax/MiniMax-M2.7
pnpm openclaw models status

If you’re in China, use:

pnpm openclaw onboard --auth-choice minimax-cn-api

#

A couple gotchas:

minimax/MiniMax-M2.7 is case-sensitive
NVIDIA’s bundled provider docs show nvidia/minimaxai/minimax-m2.5, not M2.7
if you use MiniMax OAuth instead of API key, the model ref becomes minimax-portal/MiniMax-M2.7

Docs:

MiniMax: https://docs.openclaw.ai/providers/minimax
Models CLI: https://docs.openclaw.ai/cli/models

If it throws Unknown model, paste:

pnpm openclaw --version
pnpm openclaw models list
pnpm openclaw models status

lofty obsidian Apr 13, 2026, 1:58 PM

#

https://build.nvidia.com/minimaxai/minimax-m2.7/modelcard

knotty brookBOT Apr 13, 2026, 1:59 PM

#

Ah, nice catch, then NVIDIA added it upstream.

I checked the current OpenClaw NVIDIA provider, and its bundled catalog is still behind, it still includes nvidia/minimaxai/minimax-m2.5, not M2.7.

So I’d try this:

pnpm openclaw --version
pnpm openclaw models scan
pnpm openclaw models list

If you see nvidia/minimaxai/minimax-m2.7, then set it:

pnpm openclaw models set nvidia/minimaxai/minimax-m2.7

#

If it doesn’t show up, easiest path is to use MiniMax directly instead of NVIDIA for now:

export MINIMAX_API_KEY="sk-..."
pnpm openclaw onboard --auth-choice minimax-global-api
pnpm openclaw models set minimax/MiniMax-M2.7

So short version: the NVIDIA model exists, but OpenClaw may not know it yet until the NVIDIA provider catalog catches up.

Docs:

NVIDIA: https://docs.openclaw.ai/providers/nvidia
MiniMax: https://docs.openclaw.ai/providers/minimax
Models CLI: https://docs.openclaw.ai/cli/models

If you paste the output of pnpm openclaw models list, I can tell you which exact ref OpenClaw sees.

lofty obsidian Apr 13, 2026, 2:00 PM

#

where would i add the base url and api key though?

knotty brookBOT Apr 13, 2026, 2:01 PM

#

In your OpenClaw config, usually:

pnpm openclaw config file

That prints the active config path. Default is:

~/.openclaw/openclaw.json

For NVIDIA, the base URL goes under:

models.providers.nvidia.baseUrl

and the API key is best kept in an env var:

#

NVIDIA_API_KEY

So the config looks like this, and yes, it’s JSON5, not YAML:

{
  models: {
    mode: "merge",
    providers: {
      nvidia: {
        baseUrl: "https://integrate.api.nvidia.com/v1",
        api: "openai-completions",
        apiKey: "${NVIDIA_API_KEY}",
      },
    },
  },

#

Then set the key in your shell / env:

export NVIDIA_API_KEY="nvapi-..."

If you want to check the current value:

pnpm openclaw config get models.providers.nvidia

One important caveat: current OpenClaw docs/source still show the bundled NVIDIA catalog lagging behind NVIDIA’s site, so if nvidia/minimaxai/minimax-m2.7 doesn’t resolve, that’s probably why.

#

Docs:

Config: https://docs.openclaw.ai/gateway/configuration
NVIDIA: https://docs.openclaw.ai/providers/nvidia
Config CLI: https://docs.openclaw.ai/cli/config

If you want, I can write the exact full openclaw.json snippet for NVIDIA M2.7 next.

lofty obsidian Apr 13, 2026, 2:06 PM

#

{
“agents”: {
“defaults”: {
“workspace”: “/home/neel/.openclaw/workspace”,
“models”: {
“openrouter/auto”: {
“alias”: “OpenRouter”
},
“nvidia/minimaxai/minimax-m2.7”: {}
},
“model”: {
“primary”: “nvidia/minimaxai/minimax-m2.7”
}
}
},

what can i change here?

knotty brookBOT Apr 13, 2026, 2:07 PM

#

You’d keep the agents.defaults.model.primary part, and add the provider config separately at the root under models.providers.nvidia.

Also, small but important: your pasted snippet has smart quotes (“ ”). In the real file it must use normal quotes ("), or OpenClaw won’t parse it.

Use this shape:

{
  agents: {
    defaults: {
      workspace: "/home/neel/.openclaw/workspace",
      models: {
        "openrouter/auto": { alias: "OpenRouter" },
        "nvidia/minimaxai/minimax-m2.7": {},
      },
      model: {

#

        primary: "nvidia/minimaxai/minimax-m2.7",
      },
    },
  },

  models: {
    mode: "merge",
    providers: {
      nvidia: {
        baseUrl: "https://integrate.api.nvidia.com/v1",
        api: "openai-completions",
        apiKey: "${NVIDIA_API_KEY}",
      },
    },
  },

#

Then set the key in your env:

export NVIDIA_API_KEY="nvapi-..."

After that, run:

pnpm openclaw config validate
pnpm openclaw models status

#

One caveat: if it still says unknown model, that’s probably because OpenClaw’s current NVIDIA catalog is still behind and doesn’t know minimax-m2.7 yet. In that case, the fix is not this block, it’s switching to direct MiniMax instead:

primary: "minimax/MiniMax-M2.7"

with MINIMAX_API_KEY.

If you want, paste the full config and I’ll point at the exact line to add.

lofty obsidian Apr 13, 2026, 2:18 PM

#

models.providers.nvidia.models: Invalid input: expected array, received undefined

knotty brookBOT Apr 13, 2026, 2:20 PM

#

Yep, that error means OpenClaw is treating models.providers.nvidia as a full manual provider config, and for manual providers models is required.

So a partial block like:

nvidia: {
  baseUrl: "...",
  apiKey: "...",
}

will fail validation, because it also expects:

models: [ ... ]

#

What I’d do

For NVIDIA, don’t add models.providers.nvidia manually unless you want to fully define the whole model catalog yourself.

Instead:

remove the manual models.providers.nvidia block
set the env var only:

export NVIDIA_API_KEY="nvapi-..."

The bundled NVIDIA provider auto-enables from that key.

But for your exact case

#

Since you want minimax-m2.7, and OpenClaw’s current NVIDIA catalog looks behind, the least painful fix is to use MiniMax directly instead:

export MINIMAX_API_KEY="sk-..."
pnpm openclaw onboard --auth-choice minimax-global-api
pnpm openclaw models set minimax/MiniMax-M2.7

So short version:

NVIDIA manual block → requires full models: [...]
Bundled NVIDIA → just use NVIDIA_API_KEY
MiniMax M2.7 today → better to use the MiniMax provider directly

If you want, I can paste the smallest working config for the MiniMax-direct setup.

lofty obsidian Apr 13, 2026, 2:22 PM

#

“models”: {
“providers”: {
“nvidia”: {
“baseUrl”: “https://integrate.api.nvidia.com/v1”,
“api”: “openai-completions”
}
}
},
“agents”: {
“defaults”: {
“model”: { “primary”: “nvidia/minimaxai/minimax-m2.7” }
}
},

this is how it looks for me right now

{
env: { NVIDIA_API_KEY: “nvapi-...” },
models: {
providers: {
nvidia: {
baseUrl: “https://integrate.api.nvidia.com/v1”,
api: “openai-completions”,
},
},
},
agents: {
defaults: {
model: { primary: “nvidia/nvidia/nemotron-3-super-120b-a12b” },
},
},
}

this is config exmaple from docs

knotty brookBOT Apr 13, 2026, 2:22 PM

#

Yep, your config shape is basically fine.

The important difference is this:

the docs example uses a model the bundled NVIDIA provider already knows, like
nvidia/nvidia/nemotron-3-super-120b-a12b
your config uses
nvidia/minimaxai/minimax-m2.7

So the problem is probably not baseUrl or api, it’s that minimax-m2.7 is not in OpenClaw’s current bundled NVIDIA catalog yet.

Quick sanity check

Temporarily try a known NVIDIA model:

#

pnpm openclaw models set nvidia/nvidia/nemotron-3-super-120b-a12b
pnpm openclaw models status

or:

pnpm openclaw models set nvidia/minimaxai/minimax-m2.5

If that works, then your NVIDIA auth/config is fine, and the only issue is the m2.7 model ref.

So what to do

If you want MiniMax M2.7 right now, I’d switch to the direct MiniMax provider instead:

#

export MINIMAX_API_KEY="sk-..."
pnpm openclaw onboard --auth-choice minimax-global-api
pnpm openclaw models set minimax/MiniMax-M2.7

Also, if you copied this from Discord, make sure the real file uses normal quotes, not curly quotes.

If you want, I can give you the exact before/after config for switching from NVIDIA M2.7 to direct MiniMax M2.7.

lofty obsidian Apr 13, 2026, 2:32 PM

#

i tried models listed in the doc as well, doesnt work

knotty brookBOT Apr 13, 2026, 2:32 PM

#

Got it, then it’s not the model id anymore, it’s almost certainly auth or config loading.

Please paste these exactly:

pnpm openclaw --version
pnpm openclaw config file
pnpm openclaw gateway status
pnpm openclaw models status --probe-provider nvidia
pnpm openclaw config get models.providers.nvidia

If the gateway is running as a service, also paste:

pnpm openclaw status --all

#

What I’m looking for:

wrong config file
NVIDIA_API_KEY not visible to the gateway
gateway/service not reloading after config changes
provider probe failing even though the config looks right

If you have the exact error line from models status --probe-provider nvidia, paste that too, that’ll probably give it away fast.

lofty obsidian Apr 13, 2026, 2:34 PM

#

🦞 OpenClaw 2026.4.2 (d74a122) — Your task has been queued; your dignity has been deprecated.

Config : ~/.openclaw/openclaw.json
Agent dir : ~/.openclaw/agents/main/agent
Default : nvidia/minimaxai/minimax-m2.7
Fallbacks (0) : -
Image model : -
Image fallbacks (0): -
Aliases (1) : OpenRouter -> openrouter/auto
Configured models (3): openrouter/auto, nvidia/minimaxai/minimax-m2.7, nvidia/nemotron-3-super-120b-a12b

Auth overview
Auth store : ~/.openclaw/agents/main/agent/auth-profiles.json
Shell env : off
Providers w/ OAuth/tokens (1): openai-codex (1)

nvidia effective=env:nvapi-zu...ctTfnTzH | env=nvapi-zu...ctTfnTzH | source=env: NVIDIA_API_KEY
openai-codex effective=profiles:~/.openclaw/agents/main/agent/auth-profiles.json | profiles=1 (oauth=1, token=0, api_key=0) | openai-codex:default=OAuth
openrouter effective=profiles:~/.openclaw/agents/main/agent/auth-profiles.json | profiles=1 (oauth=0, token=0, api_key=1) | openrouter:default=sk-or-v1...b152ce62

knotty brookBOT Apr 13, 2026, 2:35 PM

#

Nice, this is actually good news, your config and NVIDIA key are being seen.

So we can stop blaming openclaw.json. The next question is just:

does NVIDIA auth/probe work?
does Nemotron work?
does MiniMax M2.7 fail specifically?

Please run these and paste the output:

pnpm openclaw models status --probe-provider nvidia

pnpm openclaw infer model run --model nvidia/nvidia/nemotron-3-super-120b-a12b --prompt "say hi" --json

#

pnpm openclaw infer model run --model nvidia/minimaxai/minimax-m2.7 --prompt "say hi" --json

Why these 3:

if probe fails → auth/endpoint problem
if Nemotron works but M2.7 fails → NVIDIA/OpenClaw model-specific issue
if both fail → NVIDIA setup issue despite the key being detected

We’re pretty close now.

lofty obsidian Apr 13, 2026, 2:35 PM

#

that was the output of probe and infer command doesnt exist apparently

knotty brookBOT Apr 13, 2026, 2:36 PM

#

Ah, that’s on me, sorry. If infer isn’t exposed in your build, let’s use the older agent smoke test instead.

Please run these:

pnpm openclaw agent --model nvidia/nvidia/nemotron-3-super-120b-a12b --message "Reply with exactly: nemotron-ok"

pnpm openclaw agent --model nvidia/minimaxai/minimax-m2.7 --message "Reply with exactly: m27-ok"

If you want to sanity check the command surface too:

pnpm openclaw --help

#kimi

What I’d do

But for your exact case

Quick sanity check

So what to do