CLI · EARLY SUPPORT

Use Pi.dev with a local LLM on Mac

Try Pi.dev with a local LLM on Mac using Bridge1. Early support is a one-turn smoke test; copy the generated custom-provider JSON and review it.

Updated 2026-10-29

Pi.dev has early support: a one-turn smoke test only. Setup may change.

Two featureless clay walkers cross a stone footbridge, with an orange backpack on the trailing traveller.

Before connecting Pi.dev

For Pi.dev, start with an Apple Silicon Mac: 24 GB minimum, 32 GB or more recommended. Check memory fit.
Bridge1 lists Ornith 1.5 35B-A3B (OptiQ 4-bit · default for coding) as its default for coding. Check the model's memory and context before choosing it for Pi.dev.
Install Bridge1 through invite-only macOS early access, choose a model and open the Gateway panel for Pi.dev.

Pi.dev API and model names

Pi.dev uses OpenAI-format Chat Completions: api is openai-completions and baseUrl ends in /v1.

Set id to route1 to follow your Bridge1 selection.

How Route1 chooses a model →

Start from the terminal

bridge1 launch pi

What you skip: Bridge1 does not edit Pi’s saved settings or session folders; it loads a temporary provider extension. Pi may write its own files during a session.
Versions: Pi refuses older versions and other majors; newer minors and patches warn.
Needs: On macOS, Pi needs an available model and suitable automatic-loading settings; manual steps below still work.

Set up Pi.dev

Copy the generated JSON into ~/.pi/agent/models.json. Create the file if needed. This example shows Route1 and one onboarded model; your Gateway panel lists your own models.

Under providers.bridge1, keep baseUrl, api, apiKey and the models array.
The models array lists all onboarded models. Keep each id and the generated input, contextWindow and maxTokens fields when present; input is omitted for models without vision. Copy the JSON again after onboarding another model.

{
  "providers": {
    "bridge1": {
      "baseUrl": "http://127.0.0.1:54321/v1",
      "api": "openai-completions",
      "apiKey": "YOUR-GATEWAY-KEY",
      "models": [
        {
          "id": "route1",
          "input": [
            "text",
            "image"
          ],
          "contextWindow": 262144,
          "maxTokens": 32768
        },
        {
          "id": "mtp-qwen3.6-35b-a3b-gguf",
          "input": [
            "text",
            "image"
          ],
          "contextWindow": 262144,
          "maxTokens": 32768
        }
      ]
    }
  }
}

First connection check

In Pi.dev, type bridge1: help in a chat to see status, list models, or load, switch or stop a model. If a model is not ready, Bridge1 explains the next step in the chat. You can turn chat controls off in Settings.

Check that it works

Confirm Pi.dev uses Bridge1 and your selected model; route1 follows the Bridge1 choice. Keep generated input, contextWindow and maxTokens fields when supplied.
Ask Pi an ordinary question. A relevant answer is the published short smoke-turn check.
Only after separately authorizing a task, use a disposable project and request a greeting file. Inspect its contents; this optional check extends beyond published smoke coverage.

A Pi model list, connection test or echoed reply does not prove a completed task.

If something is off

Symptom	What it usually means	What to check	More help
Wrong provider or missing fields	The manual entry may be incomplete.	Recopy the generated Bridge1 provider JSON.	Pi configuration
Launcher warns about the version	The installed version is untested.	Compare the rule below.	Pi support
Model unavailable	Loading depends on availability and settings.	Check the selected model in Bridge1.	Model availability

Versions

Bridge1 0.7.7 launches Pi.dev 1.4.0, tested on macOS on 2026-10-27. This check used bridge1 launch pi. Below the tested version or a different major version: refused. A newer minor of the same major proceeds with a one-line warning. A newer patch warns. These are not the only versions that work; you do not need the tested version. The manual setup below was not re-tested at that version.

Pi.dev: limits to check

Pi.dev has had one short smoke-test turn. Check the generated configuration before relying on it.

Review Pi.dev and shared client limits →

Other client setup guides

Request Bridge1 early access

Pi.dev setup questions

How much testing has Pi.dev had with Bridge1?

Pi.dev has early support: one short smoke-test turn. This does not establish end-to-end session testing.

What does the generated Pi.dev provider contain?

Its bridge1 provider contains baseUrl, api, apiKey and models. Copy the Gateway panel JSON for all onboarded models. Copy it again after onboarding another model.

Why keep the Pi.dev contextWindow value?

Without contextWindow, Pi assumes 128,000 tokens. For Route1, Bridge1 generates the smallest window among its possible models.