Updated 2026-06-16

DeepSeek nanobot setup: use the official config.json contract instead of piecing together random agent examples

DeepSeek's official nanobot integration page is useful because it describes the exact setup path instead of leaving users to infer how another agent should talk to DeepSeek. The page installs nanobot with `uv`, initializes a config file with `nanobot onboard`, then shows the concrete `config.json` fields that matter: model, provider, API key, and `apiBase`. That makes this a strong support-query page for developers searching how to wire nanobot to DeepSeek V4 without endpoint confusion.

1. What nanobot is in DeepSeek's official docs

DeepSeek describes nanobot as a lightweight AI agent that supports integration with popular chat tools. That framing matters because this is an agent-connectivity page, not a generic DeepSeek chat tutorial.

The right search intent here is setup discipline: install path, config location, provider choice, and the exact DeepSeek endpoint values that the official page publishes.

Sources checked

2. Install nanobot with uv before editing config

DeepSeek's guide starts by requiring `uv`, then installs nanobot with `uv tool install nanobot-ai`. That runtime assumption is the first thing to validate on a fresh machine.

The same page also documents Windows PATH cleanup and `uv tool update-shell`, which is useful context if a teammate says the install succeeded but the command is still missing.

uv tool install nanobot-ai
nanobot --version

3. `nanobot onboard` creates the config baseline

DeepSeek's setup flow does not jump directly into manual file creation. It says to run `nanobot onboard` first so the initial configuration file exists in the expected OS-specific location.

On Linux and macOS, the file lives at `~/.nanobot/config.json`. On Windows, it lives under the user's home directory. That matters because wrong-path edits are a common reason agent configs appear to be ignored.

nanobot onboard

4. The official DeepSeek provider wiring in config.json

DeepSeek's page shows a simple but important configuration shape: `agents.defaults.model` set to `deepseek-v4-pro`, `agents.defaults.provider` set to `deepseek`, and a matching provider block with the API key and `apiBase`.

The official `apiBase` is `https://api.deepseek.com/v1`, which makes this page especially useful for users who are mixing OpenAI-format and Anthropic-format examples from unrelated tools.

{
  "agents": {
    "defaults": {
      "model": "deepseek-v4-pro",
      "provider": "deepseek"
    }
  },
  "providers": {
    "deepseek": {
      "apiKey": "<your DeepSeek API Key>",
      "apiBase": "https://api.deepseek.com/v1"
    }
  }
}

Sources checked

5. Why this page is Pro-first and how to think about Flash

DeepSeek's official example uses `deepseek-v4-pro`, so that is the known-good baseline. Start there when you are proving the integration works.

If your nanobot workflow becomes broader or more cost-sensitive, you can benchmark `deepseek-v4-flash` later. Keep the provider and `apiBase` unchanged while changing only the model, so you know whether the difference is cost-quality routing rather than broken endpoint setup.

6. Start the agent and debug in the right order

DeepSeek's page starts nanobot with `nanobot agent`. If the session fails, verify the config path, the provider name, the API key field, and the `apiBase` value before rewriting prompts or blaming the model.

This order matters because most nanobot setup problems are ordinary configuration errors. A wrong provider label or endpoint string will look like a model problem until you inspect the file carefully.

nanobot agent

FAQ

How do I install nanobot for DeepSeek?

DeepSeek's official guide installs it with `uv tool install nanobot-ai`, then checks the install with `nanobot --version`.

Which file should I edit for DeepSeek nanobot setup?

DeepSeek's page says the config lives at `~/.nanobot/config.json` on Linux and macOS after you run `nanobot onboard`.

Which base URL should the DeepSeek provider use in nanobot?

Use `https://api.deepseek.com/v1`, which is the `apiBase` value shown on the official DeepSeek nanobot page.

What model does the official example use?

The documented baseline is `deepseek-v4-pro`.

Can I switch nanobot to Flash later?

Yes. Once the baseline integration works, you can benchmark `deepseek-v4-flash` as a cheaper route while keeping the same provider and endpoint wiring.

The official nanobot page is valuable because it removes ambiguity: `uv` install, `nanobot onboard`, a known config path, a named `deepseek` provider, the `/v1` base URL, and a Pro-first model choice. Start with that exact contract, then optimize only after the baseline agent session is stable.

Related model comparisons

Continue from this guide into structured DeepSeek-first comparison pages with model tables, routing advice, and pricing context.