Integrations

Plug 1,700+ uncensored models into anything.

DemonRoute is an OpenAI-compatible API. If your app can talk to OpenAI, it can talk to us — same requests, different base URL and model names. This is the general guide; per-tool walkthroughs land here as each partnership ships.

Get a key → Browse the catalog

Featured integration Made by DemonRoute

Demon Build — describe an app, get a working website.

Our own builder rides this exact API. Prompt it in plain English and it designs, codes, and ships a real multi-page site — landing pages, dashboards, tools, stores — then iterates with you in chat. Every model in the DemonRoute catalog is available to power your builds, and it bills straight to your existing DemonRoute balance. One key, two products.

▸Full-stack sites from a single prompt
▸Chat-driven edits and rebuilds
▸Pick any catalog model as the builder brain
▸Uses your DemonRoute key & balance — nothing new to buy

Open Demon Build → Launch the builder

Guide 01

General — any application

Current

DeepChat · Fabric · SillyTavern · tgpt · Twinny · PR-Agent · RAGFlow · FreeCAD AI · OmniRoute · BlindPilot · incognide · RagLeap · llm-subtrans · Seren live now

Get an API key

Two ways in: standard — sign up with an email at demonroute.com — or anonymous: pay with crypto, no email, no IP logs. Either way you get a key. Keep it secret.

Point your app at the base URL

Every request goes to one endpoint. Authentication is a bearer token:

# base URL
https://api.demonroute.com/v1

# header
Authorization: Bearer YOUR_KEY

Send your first request

Same shape as the OpenAI chat completions API — pick any model ID from the catalog:

# curl
curl https://api.demonroute.com/v1/chat/completions \
  -H "Authorization: Bearer $DEMONROUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "dphn/Dolphin-Mistral-24B-Venice-Edition",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

# Python — official openai package
from openai import OpenAI

client = OpenAI(
    base_url="https://api.demonroute.com/v1",
    api_key="YOUR_KEY",
)

resp = client.chat.completions.create(
    model="dphn/Dolphin-Mistral-24B-Venice-Edition",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

// Node — official openai package
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.demonroute.com/v1",
  apiKey: process.env.DEMONROUTE_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "dphn/Dolphin-Mistral-24B-Venice-Edition",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(resp.choices[0].message.content);

That's it. Any OpenAI-compatible client, SDK, or tool works — swap the base URL and key and you're running on DemonRoute.

Model IDs

Model IDs look like family/Name — for example dphn/Dolphin-Mistral-24B-Venice-Edition or XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. The ID is the exact string you pass as model. Copy it straight from the catalog — no translation, no prefixes. You can also list every live model from the API:

curl https://api.demonroute.com/v1/models \
  -H "Authorization: Bearer $DEMONROUTE_API_KEY"

Pricing

Every model shows its price per 1M tokens (input / output) on the catalog page and pricing page. What you see is what you pay — prepaid balance, no tiers, no surprises. A few models carry a free lane at $0.00.

Guide 02

LiteLLM — proxy & SDK

Tested against the live gateway

Completion · streaming · proxy

Install LiteLLM

DemonRoute works through LiteLLM's standard openai/ provider prefix — the supported path for any OpenAI-compatible endpoint. No plugin, no fork.

pip install "litellm[proxy]"

Run the proxy

One config.yaml entry per model you want behind the proxy. model_name is what your apps ask for; model is the catalog ID with the openai/ prefix:

# config.yaml
model_list:
  - model_name: dolphin-venice
    litellm_params:
      model: openai/dphn/Dolphin-Mistral-24B-Venice-Edition
      api_base: https://api.demonroute.com/v1
      api_key: os.environ/DEMONROUTE_API_KEY

# LiteLLM requires a master key for the proxy itself
export LITELLM_MASTER_KEY=sk-anything-you-pick
litellm --config config.yaml --port 4000

Call it like OpenAI

The proxy now speaks OpenAI on localhost:4000 — every LiteLLM feature (fallbacks, rate limits, spend tracking, logging) sits in front of DemonRoute:

curl http://localhost:4000/v1/chat/completions \
  -H "Authorization: Bearer sk-anything-you-pick" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "dolphin-venice",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Or skip the proxy — SDK direct

If you only need completion() from Python, no proxy required. Streaming works the same way:

from litellm import completion

resp = completion(
    model="openai/dphn/Dolphin-Mistral-24B-Venice-Edition",
    api_base="https://api.demonroute.com/v1",
    api_key=os.environ["DEMONROUTE_API_KEY"],
    messages=[{"role": "user", "content": "Hello"}],
)

Verified live against this gateway: completion, streaming, and the full proxy round-trip all pass. Note — LiteLLM's newer openai_like/ provider is broken in current releases (1.104.x); use the openai/ prefix shown here.

Integrations

Works with the tools you already run

Because the API is OpenAI-compatible, it drops into anything that accepts a custom base URL. These apps ship DemonRoute as a native provider — or are about to. The pattern is always the same: base URL + key + model ID.

Live now — merged upstream

pi-openai-compat ↗

Works now

An extension for the pi coding agent that registers OpenAI-compatible providers as native models in pi's model picker. DemonRoute is a first-class provider template: run /compat-login, pick Demon Route, paste your key, and code with uncensored open-weight models.

Merged — PR #43 — in main now

/compat-login → Demon Route → key

Postmill AI ↗

Works now

An AI app with team organizations and a built-in hub. DemonRoute is an LLM provider in Settings (Beta/v1): pick it, paste your key, and run uncensored open-weight models and persona fine-tunes across text, tools, and the hub.

Merged — issue #191 — deployed in production

Settings → LLM Providers → DemonRoute → key

DevoxxGenie ↗

Works now

An LLM plugin for JetBrains IDEs — AI chat and agent mode inside IntelliJ. DemonRoute is a built-in router definition next to OpenRouter: enable it in the Routers settings tab, paste your key, and run uncensored models on your code. The maintainer built it into his new generic router provider after our PR opened the conversation.

Merged — PR #1384 — implemented by the maintainer

Settings → Routers → DemonRoute → key

ModelDeck ↗

Works now

A multi-provider BYOK chatbot testing playground and benchmarking suite. DemonRoute is a built-in provider, next to OpenRouter: pick it, paste your key, and test uncensored open-weight models side by side.

Merged — PR #1

Inkwell ↗

Works now

A desktop Markdown viewer in Rust with a built-in bring-your-own-key LLM copilot. DemonRoute is a built-in provider preset: pick it in Settings, paste your key, and run uncensored models against your documents.

Merged — PR #14 — shipped in v0.4.0

Provider presets → DemonRoute → key

llm-subtrans ↗

Works now

LLM-powered subtitle translator — translate entire subtitle files with AI. DemonRoute is a built-in gateway provider: set your DEMONROUTE_API_KEY, pick it from the provider list, and translate with uncensored models that handle the dialogue the big APIs won't touch.

Merged — PR #530 — ships in their next release

RagLeap ↗

Works now

Open-source RAG platform for business owners — chat with your documents over any LLM. DemonRoute is a first-class provider: set LLM_PROVIDER=demonroute, paste your key, pick any model. Prepaid with a hard stop at $0, so an AI employee answering questions overnight can't create a surprise invoice.

Merged — PR #851

incognide ↗

Works now

Desktop AI app built around "own your data" — local models plus first-class cloud providers. DemonRoute is a built-in cloud provider next to OpenRouter and OrcaRouter: pick it from the provider list, paste your key, no config-file hand-editing.

Merged — PR #371 — ships in their next release

BlindPilot ↗

Works now

An accessible desktop frontend for AI agents and chat. DemonRoute is a built-in named provider: pick it from the provider list, paste your key — it's stored in the system password store — and the model list loads from the endpoint on its own. No base URL to type.

Works now — built-in since v0.47.14

Chat → Accounts → DemonRoute → key

Seren ↗

Works now

An AI-powered IELTS speaking coach with Cambridge-style diagnostic tests, CEFR evaluation, personalized lessons, and real-time voice practice. DemonRoute is a built-in provider: open Settings, paste your key, and choose a model or leave it on auto to discover models from the endpoint.

Works now — built-in provider in master

Settings → DemonRoute → API key

OmniRoute ↗

Works now

Free MIT AI gateway: one endpoint routing to 359 providers and 1200+ models, working with Claude Code, Codex, Cursor, Cline and Copilot. DemonRoute landed as a built-in provider — uncensored models with live catalog and embeddings, available across every coding tool the gateway serves.

Merged — PR #17001 — ships in v3.8.52

Twinny ↗

Works now

A VS Code extension for AI chat, code review, and agent mode — our first VS Code integration. DemonRoute ships as a hosted provider preset: pick it from the provider list, paste your key, and run uncensored models on your code without leaving your editor.

Merged — PR #545 — ships in their next release

Provider presets → DemonRoute → key

LiraVtuber ↗

Works now

An AI VTuber with a control panel and WhatsApp bridge. DemonRoute is a built-in opt-in provider: set LLM_PROVIDER=demonroute, paste your key, and run uncensored character models — Cydonia by default, with Hermes and abliterated fallbacks and key rotation built in.

Implemented by the maintainer — issue #1 — commit 801fe72

RAGFlow ↗

Works now

An open-source RAG engine for document search and Q&A. DemonRoute is a supported model provider, so your pipelines can run on uncensored models.

Merged — PR #20853 — ships in their next release

FreeCAD AI ↗

Works now

An AI workbench for FreeCAD. DemonRoute is a provider preset — run uncensored models against your CAD workflows from inside FreeCAD.

Merged — PR #129 — ships with their next release

Provider presets → DemonRoute → key

DeepChat ↗

Works now

A polished desktop AI chat client with agents, MCP tools, and skills. DemonRoute ships as a built-in provider preset: select it from the provider list, paste your key, and chat with uncensored models in a real app instead of a terminal.

Beta channel — v1.1.3-beta.3 and up

Provider list → DemonRoute → key

Fabric ↗

Works now

A CLI that runs reusable AI prompt patterns over anything you pipe into it. DemonRoute is a built-in vendor: select it, paste your key, and run uncensored models through Fabric's pattern library.

Merged — rolling out to package managers

vendor: DemonRoute

Agenta ↗

Works now

Open-source LLMOps platform — build, evaluate, and observe LLM apps and agents with your team. DemonRoute is named in Agenta's provider docs alongside OpenAI, Anthropic, and OpenRouter: connect it as an OpenAI-compatible endpoint, paste your key, and run uncensored open-weight models through the same playground and evals.

Merged — PR #7508 — ships with v0.123.0

SillyTavern ↗

Works now

The power-user frontend for character and roleplay chat. Point it at DemonRoute as a custom OpenAI-compatible endpoint and run uncensored models behind your characters.

Official docs walkthrough landing in their next release · our step-by-step below ↓

Custom (OpenAI-compatible) → base URL + key

SillyTavern · full walkthrough

Create your key

Sign up at demonroute.com (standard or anonymous — both work), open Console → Keys, and create a key. Copy it somewhere safe. Any deposit, even $1, unlocks the complimentary lane: 50 free requests every day.

Open API connections

In SillyTavern, click the plug icon (API Connections) at the right end of the top bar.

Select the source

Set API to Chat Completion, then Source to Custom (OpenAI-compatible).

Point it at DemonRoute

Fill in the two fields and press Connect:

Custom Endpoint (Base URL): https://api.demonroute.com/v1
API Key: sk-…your key…

The /v1 matters, and no trailing slash. If Connect fails, that is almost always it.

Pick a model

The model dropdown now lists the full DemonRoute catalog — 170-plus roleplay-tagged models. There is no wrong answer, but if you want a starting point, these are community favorites (suggestions only, not endorsements — taste is personal and the catalog page has them all with prices):

Sao10K/L3-8B-Lunaris-v1 · 8B · $0.37 / $0.38 per M tokens
TheDrummer/Cydonia-R1-24B-v4.1 · 24B · $2.80 / $4.64 per M
ArliAI/Llama-3.3-70B-ArliAI-RPMax-v1.4 · 70B · $10.40 / $12.00 per M

Rough guide: 8B is fast and cheap and surprisingly good, 24B is the sweet spot most roleplayers land on, 70B is for long character cards and heavy prose. Prices are per million tokens in/out — a typical chat message costs a fraction of a cent on the small models. Everything is prepaid, so you can never overspend.

Tune for characters

Raise the context slider toward the model's listed context so it remembers more of the chat, and if replies cut off mid-sentence raise the response length. Streaming is on by default and just works. Failed calls cost nothing — they refund automatically within five minutes.

Troubleshooting

Error: “This model does not accept a system message after an assistant message.” Some upstreams insist system messages sit at the very start of the chat, while SillyTavern sprinkles them throughout. Fix: open the AI Response Configuration panel (the sliders icon) and tick Squash system messages — or set Prompt Post-Processing to Merge consecutive messages. One click and the error is gone.

Connect fails or no models in the dropdown? Check the endpoint is exactly https://api.demonroute.com/v1 — the /v1 matters, no trailing slash — and that the key pasted cleanly with no spaces.

Slow first reply, or an empty one? Some models can take 30-60 seconds to produce their first token, especially when they are warming up - give them a moment before hitting stop. Some models interacting with SillyTavern also have a higher likelihood of producing no output on the first attempt for certain prompts. No harm done: any call that does not produce an output is refunded automatically within five minutes, and the same prompt usually goes through when retried - just swipe or send again.

Anything else: failed calls refund automatically within five minutes, so experimenting costs nothing.

iflashc ↗

Works now

A flashcard app that turns new words into study cards with definitions and translations into your native language. DemonRoute is a built-in translator option: pick demonroute, paste your key, and translate with any of 1,300+ models — most of them bilingual.

Merged — PR #17 — in master now

tgpt / Terminal GPT ↗

Works now

A cross-platform terminal chatbot. The DemonRoute provider adds one-key access to the full catalog from your terminal.

Merged — shipping in their next release

Landing soon — committed, live in their next release

CreativeWriter ↗

Committed

A novel-writing studio with AI rewrite, character chat, and story codex tools. DemonRoute will be listed as an OpenAI-compatible provider option, so writers can run uncensored models on their drafts.

Shipping in their next release

Universal — anything that speaks OpenAI

Python / Node SDKs

Works now

The official openai packages. Set the base URL, keep the rest of your code.

base_url="https://api.demonroute.com/v1"

LiteLLM

Works now

Run DemonRoute as a provider in your LiteLLM proxy — tested config in Guide 02 above.

api_base: https://api.demonroute.com/v1

Your own app

Works now

Any HTTP client that can POST JSON. Streaming and non-streaming chat completions.

POST /v1/chat/completions

Per-tool guides land here as partnerships ship.

This page is the general pattern. Dedicated walkthroughs — screenshots, exact settings, ready-made configs — get added for each integration we formalize. If you maintain a tool and want DemonRoute as a first-class provider, email jeremy@nftdemon.com — partnerships welcome.

Read the docs Check live status