AI Badgr integrations

40 frameworks, model servers and developer tools. Each page shows how to start, what it needs, what it costs and how well the path is tested.

Six ways to use Badgr

Badgr runs, hosts and automates AI software without you managing the machines underneath. Use the dashboard, the CLI, the API, an SDK or the tools you already have.

  1. 1.Deploy an application or model

    Pick a model or a supported app template. Badgr finds suitable compute, starts it and returns an address to call or open.

    Interface: Serve. You receive: A model address or an app URL.

    Read the docs

  2. 2.Run a workload

    Give Badgr a ComfyUI workflow, a Python script or a training job. Badgr runs it on CPU or GPU, streams logs and stops when the work finishes.

    Interface: Run. You receive: Outputs, logs and a receipt.

    Read the docs

  3. 3.Execute a coding agent

    Assign Codex or Claude Code a task. The agent works in an isolated environment, apart from your machine.

    Interface: Launch. You receive: A patch, test results and artifacts.

    Read the docs

  4. 4.Call AI models

    Send requests from an SDK or plain HTTP. Badgr answers with a model response. This path stays separate from jobs and services.

    Interface: AI API. You receive: A model response.

    Read the docs

  5. 5.Connect an existing framework

    Use LangGraph, LiteLLM and similar tools you already have. They call Badgr as their model provider, or call it for compute.

    Interface: Any interface. You receive: Model responses or job results.

    See the pages

  6. 6.Automate with CI

    Trigger Badgr from GitHub Actions or Azure DevOps. A pipeline step starts the work, checks its status and reads results. Each step calls Run, Serve or Launch once.

    Interface: Run, Serve or Launch. You receive: Status, callbacks and artifacts.

    Read the docs

How it fits together

Six ways in, four interfaces, three lifecycles and one shared platform. Badgr does not replace the frameworks, models and agents you already use. It runs them, hosts them or supplies them with compute.

Four interfaces and three lifecycles

AI API

Lifecycle: Invoke

Request, then response. Connect models.

Run (Jobs)

Lifecycle: Job

Finite CPU or GPU work. Run workloads.

Serve (Deploy)

Lifecycle: Service

Software that stays available. Host models and apps.

Launch (Agents)

Lifecycle: Job

A job with agent checks. Execute agents.

  • Invoke

    Request, then response

    Used by AI API

  • Job

    Start, run, finish

    Used by Run and Launch

  • Service

    Deploy, ready, stop

    Used by Serve

One shared control plane

  • Identity and permissions
  • Durable execution state
  • Budgets and cost limits
  • Scheduling and placement
  • Recovery and lifecycle
  • Events, receipts and billing

Software that runs on it, and where

  • Agent sandboxes

    Codex, Claude Code, Cline

  • GPU workloads

    ComfyUI, PyTorch, training

  • Model servers

    vLLM, Ollama, SGLang

  • Applications

    Open WebUI, Dify, OpenClaw

Compute: Cloud CPU, Cloud GPU, Your own GPUs (bring your own).

Results: Model responses, images, videos, pull requests, test results, addresses, checkpoints, artifacts, logs and receipts.

Reference: compute API docs.

Who uses Badgr this way

  • Connect

    Developers who already have an application, framework or gateway.

    Your app or framework, then the Badgr API, then a model response or a result.

    They want models or compute inside existing applications without managing infrastructure.

  • Run

    Developers and creators with scripts, training configs or media workflows.

    Your workload, then Badgr Run on a GPU or CPU, then outputs, checkpoints and a receipt.

    They need temporary GPUs, repeatable runs, logs and cost limits, not an always-on server.

  • Deploy

    Developers hosting models, inference servers or compatible AI applications.

    Your model or app, then Badgr Serve, then a live address or app URL.

    They need software to stay available, with a managed lifecycle and billing.

  • Launch and agents

    Developers automating coding tasks, browsing and multi-agent work.

    A task, then Badgr Launch in an isolated sandbox, then a patch, tests, artifacts and a receipt.

    They need isolated, repeatable and bounded agent runs instead of running everything on a laptop.

All 40 integrations

Some projects sit above Badgr and call it. Others run on Badgr. A few do both.

Showing 40 integration pages. The interface shown is the main one. Many projects can use more than one.

#IntegrationInterfaceRoleRelation to BadgrWhy Badgr
1LiteLLMAI APIExternal gatewayCalls BadgrRoute model requests to Badgr
2Vercel AI SDKAI APIExternal SDKCalls BadgrInference for TypeScript apps
3One APIAI APIExternal gatewayCalls BadgrModel upstream
4LangChainAI APIExternal frameworkCalls BadgrModel calls and compute tools
5LangGraphRunExternal orchestratorCalls BadgrSubmit agent compute tasks
6PortkeyAI APIExternal gatewayCalls BadgrInference routing
7Open WebUIServeHosted app / API clientCalls BadgrHost UI or connect models
8OpenAI Python SDKAI APIExternal SDKCalls BadgrCompatible Python inference
9OpenAI Node SDKAI APIExternal SDKCalls BadgrCompatible Node inference
10Hugging Face HubRunModel sourceRuns on BadgrExecute or serve models
11PyTorchRunGPU runtimeRuns on BadgrTraining and inference
12ComfyUIRunGPU runtimeRuns on BadgrImage and video workflows
13RayRunCompute frameworkRuns on BadgrCompatible Ray jobs
14DiffusersRunGPU runtimeRuns on BadgrMedia generation
15TransformersRunModel frameworkRuns on BadgrInference and training
16KohyaRunTraining runtimeRuns on BadgrLoRA fine-tuning
17InvokeAIServeApplication runtimeRuns on BadgrAI image application hosting
18AI ToolkitRunTraining runtimeRuns on BadgrModel fine-tuning
19SimpleTunerRunTraining runtimeRuns on BadgrDiffusion training
20UnslothRunTraining runtimeRuns on BadgrEfficient LLM fine-tuning
21vLLMServeModel serverRuns on BadgrPersistent inference endpoint
22SGLangServeModel serverRuns on BadgrCompatible inference service
23OllamaServeModel serverRuns on BadgrCloud and private model hosting
24llama.cppServeModel serverRuns on BadgrGGUF model endpoint
25TensorRT-LLMServeModel runtimeRuns on BadgrOptimised NVIDIA inference
26FastChatServeServing applicationRuns on BadgrPersistent chat serving
27Triton Inference ServerServeInference serverRuns on BadgrHost model repositories
28KTransformersServeModel runtimeRuns on BadgrSpecialised CPU and GPU inference
29AIBrixServeKubernetes integrationGuide onlyKubernetes serving: guide only, not offered
30LocalAIServeModel serverRuns on BadgrHosted compatible API
31CodexLaunchAgent sandboxCalls and runsCode tasks and test verification
32ClineLaunchCoding agentRuns on BadgrRemote coding execution
33DifyAI APIExternal appCalls BadgrModel and compute integrations
34CrewAIRunExternal orchestratorCalls BadgrAgent compute tools
35OpenHandsLaunchAgent runtimeRuns on BadgrSoftware-agent tasks, started locally with one command
36AutoGenAI APIExternal orchestratorCalls BadgrModels and execution tools
37ContinueAI APIIDE clientCalls BadgrAI models in your IDE
38Browser UseLaunchBrowser agent runtimeRuns on BadgrRemote browser tasks
39Claude CodeLaunchAgent sandboxRuns on BadgrCode tasks and artifacts
40Microsoft Agent FrameworkRunExternal frameworkCalls BadgrEnterprise agent compute

Connect

Point an app or framework you already use at Badgr models.

  • LiteLLM

    Use Badgr as a model provider in LiteLLM

    documented

  • Vercel AI SDK

    Build AI apps with Vercel AI SDK and Badgr

    documented

  • One API

    Connect One API to Badgr

    documented

  • LangChain

    Run LangChain applications with Badgr

    documented

  • LangGraph

    Give LangGraph agents cloud GPU jobs

    experimental

  • Portkey

    Route model traffic through Portkey to Badgr

    experimental

  • Open WebUI

    Connect Open WebUI to Badgr models

    documented

  • OpenAI Python SDK

    Use Badgr with the OpenAI Python SDK

    documented

  • OpenAI Node SDK

    Use Badgr with the OpenAI Node SDK

    documented

  • Hugging Face Hub

    Serve supported Hugging Face models on Badgr

    documented

Run

Run your existing GPU code as a job that finishes.

  • PyTorch

    Run PyTorch training on cloud GPUs

    documented

  • ComfyUI

    Run ComfyUI workflows on demand

    documented

  • Ray

    Run single-node Ray workloads on Badgr

    experimental

  • Diffusers

    Generate images and videos with Diffusers

    documented

  • Transformers

    Run Transformers models on GPUs

    documented

  • Kohya

    Run LoRA training with Kohya

    documented

  • InvokeAI

    Run InvokeAI on a cloud GPU

    documented

  • AI Toolkit

    Run AI Toolkit fine-tuning jobs

    experimental

  • SimpleTuner

    Train diffusion models with SimpleTuner on Badgr GPUs

    experimental

  • Unsloth

    Fine-tune LLMs with Unsloth

    documented

Deploy

Start a model service and keep it available.

  • vLLM

    Serve a model with vLLM

    documented

  • SGLang

    Serve compatible models with SGLang

    documented

  • Ollama

    Run Ollama on your computer or in the cloud

    documented

  • llama.cpp

    Serve GGUF models with llama.cpp

    documented

  • TensorRT-LLM

    TensorRT-LLM on Badgr: custom container guide

    experimental

  • FastChat

    FastChat on Badgr: custom setup guide

    experimental

  • Triton Inference Server

    Triton Inference Server on Badgr: custom container guide

    experimental

  • KTransformers

    KTransformers on Badgr: compatibility guide

    experimental

  • AIBrix

    AIBrix and Badgr: where each one fits

    experimental

  • LocalAI

    LocalAI on Badgr: custom container guide

    experimental

Launch and agents

Send a task to an agent, or give an agent Badgr tools.

  • Codex

    Run Codex tasks in an isolated environment

    documented

  • Cline

    Run Cline tasks on Badgr

    documented

  • Dify

    Connect Dify applications to Badgr

    experimental

  • CrewAI

    Run compute-heavy CrewAI tasks on Badgr

    experimental

  • OpenHands

    Run OpenHands on your own computer with Badgr

    documented

  • AutoGen

    Connect AutoGen workflows to Badgr

    documented

  • Continue

    Connect Continue to Badgr models

    documented

  • Browser Use

    Run browser-use tasks on Badgr: setup guide

    experimental

  • Claude Code

    Run Claude Code remotely

    documented

  • Microsoft Agent Framework

    Add Badgr to Microsoft Agent Framework agents

    experimental

Reference docs: AI API, Jobs, Serve, Agent Cloud.