Using a Shared AI Gateway with Askimo (LiteLLM Example)

An AI gateway is an OpenAI-compatible proxy that sits in front of your AI providers and routes all team traffic through one controlled endpoint. Combined with Askimo’s built-in OpenAI-Compatible provider, it lets an entire team share one set of API keys, with no one ever seeing the real credentials.

This guide uses LiteLLM as a worked example — it is a widely-used open-source gateway that works with Askimo out of the box. Other OpenAI-compatible gateways follow the same connection pattern.

Full walkthrough available

This page is a quick reference. For a complete step-by-step setup guide using LiteLLM as the gateway, see the blog post: LiteLLM Proxy for Teams: Share ChatGPT, Claude & Gemini Without Per-User API Keys →

Why Use an AI Gateway with Askimo?

Askimo is a bring-your-own-key desktop app. Each user normally manages their own API keys locally. This is ideal for individual privacy and cost control, but creates friction for teams:

• Every member needs their own provider accounts
• Spending is fragmented across many keys
• New members can’t start until they’ve completed key setup

A shared AI gateway solves this by:

• Holding the real API keys on a server only the admin accesses
• Giving each team member a virtual key with optional spend limits
• Routing all provider traffic through one controlled endpoint
• Letting users connect Askimo with just a URL, with no provider accounts needed

Is this the right setup for your team?

Askimo + LiteLLM is a great permanent fit if your team wants centralised API key management without storing credentials on every machine, and you don’t need shared conversation history or collaboration features.

If your team needs shared workspaces, shared knowledge bases, and centralised AI governance at scale, Askimo Team is designed for that (currently in development).

Architecture

The diagram below shows how Askimo connects to a shared AI gateway (LiteLLM used as the example):

flowchart LR
    A1["Askimo Desktop\nUser 1"] --> P
    A2["Askimo Desktop\nUser 2"] --> P
    A3["Askimo Desktop\nUser 3"] --> P

    subgraph P["LiteLLM Proxy (self-hosted)"]
        direction TB
        VK["Virtual Keys\nper user"]
        RT[Model Router]
        VK --> RT
    end

    P --> OAI["OpenAI\nGPT models"]
    P --> ANT["Anthropic\nClaude models"]
    P --> GEM["Google\nGemini models"]
    P --> OLL["Ollama\nLocal models"]

LiteLLM exposes an OpenAI-style /v1 endpoint. Askimo connects to it via the OpenAI-Compatible provider, with no special plugin or configuration needed on the client side. Any gateway that exposes an OpenAI-style /v1 endpoint works the same way.

Quick-Start: Connect Askimo to a Gateway

Once your AI gateway is running (see the full guide for a LiteLLM-based setup), each team member configures Askimo as follows:

1. Open Askimo and click the provider pill in the footer bar, then click Add Provider. Or go to Settings → AI Providers.

2. Select Other providers from the list.

3. Fill in the connection details:

   Field	Value
   Provider Name	Any label (e.g. Team AI)
   Base URL	http://your-server-ip:4000/v1
   API Key	Your personal virtual key (or leave blank for open internal access)
   API Mode	Chat Completions (default)
   HTTP Version	HTTP/1.1 (default)

4. Save. Askimo calls /v1/models and populates the model list automatically.

5. Start chatting. All requests route through LiteLLM to whichever provider the model belongs to.

Include /v1 in the URL

The Base URL must include the full API path: http://192.168.1.100:4000/v1 ✅, not http://192.168.1.100:4000 ❌

What Users Get

	Without LiteLLM	With LiteLLM
API key setup per user	Required for each provider	Not needed, just one URL
Provider accounts	Each user signs up	Admin only
Spend control	Per-user, fragmented	Central, per-virtual-key budgets
Access revocation	Rotate real key (affects everyone)	Delete one virtual key
Models available	Whatever the user configured	All models the admin has enabled
Privacy	Chats stay local	Chats stay local; keys stay on server

LiteLLM Setup Summary

LiteLLM is a straightforward open-source option to get a gateway running quickly:

Step	What you do
1. Run LiteLLM	curl -sSL https://docs.litellm.ai/docker-compose.yml | docker compose --env-file .env -f - up -d
2. Add your provider API keys	Via the .env file or the Admin UI at http://localhost:4000/ui
3. Create virtual keys	One per team member, with optional spend limits and model restrictions
4. Share the proxy URL + virtual key	Each user enters it in Askimo as an OpenAI-Compatible provider

Adding Local Models (Ollama)

LiteLLM can also route to an Ollama server running on the same machine, making free open-source models available to the whole team alongside cloud models:

model_list:
  - model_name: llama3.3
    litellm_params:
      model: ollama/llama3.3
      api_base: http://host.docker.internal:11434   # use host.docker.internal when LiteLLM runs in Docker

Once added, llama3.3 appears in Askimo’s model picker alongside GPT and Claude, at no cost per request.

Troubleshooting

No models showing after connecting?

• Confirm the Base URL ends with /v1
• Check LiteLLM is running: curl http://your-server:4000/health
• Verify the virtual key has at least one model assigned

“Invalid API key” error?

• Re-check the virtual key in LiteLLM’s Admin UI
• Ensure no extra spaces when pasting the key into Askimo
• If not using virtual keys, leave the API Key field blank

Model returns errors?

• The real provider API key may be invalid or out of credits
• Check LiteLLM logs: docker logs litellm-proxy

Learn More

Full Setup Guide (Blog)

Complete walkthrough: Docker Compose, virtual keys, Ollama, remote access, and Caddy HTTPS.

Read the blog post →

OpenAI-Compatible Provider

Reference docs for connecting any OpenAI-style endpoint to Askimo.

OpenAI-Compatible provider →

Askimo Pricing & Team Edition

Team Edition brings native shared workspaces and management, with no proxy needed.

See what’s coming →

LiteLLM Documentation

Official LiteLLM proxy docs: model routing, virtual keys, budgets, and Admin UI.

docs.litellm.ai →