Skip to content
appsgit

6 jobs · 22 apps

AI for banking: open source tools you can self-host

For banks and fintech companies, the core of a private AI stack is Ollama for running open models locally, LibreChat for a private chat interface and RAGFlow for answering questions over your documents, all running on your own servers so customer and transaction data can stay on your infrastructure. Each pick was chosen by hand for its job and is shown with GitHub data from October 2026.

Financial data is often the hardest data to send to a cloud AI service. Running models and AI tools on your own servers keeps prompts, customer records and documents inside your own controls.

Updated · data from GitHub, refreshed nightly

Top picks

  1. Rank 1:

    Ollama

    ollama/ollama

    Get up and running with Llama 3.3, DeepSeek-R1, Phi-4, Gemma 3, and other large language models.

    AI & LLM ToolsGoDocker

    AI-nativeOllama has 182,289 GitHub stars, uses the MIT license, ships a Docker image and had 1,083 commits in the last year; it replaces ChatGPT.

  2. Rank 2:

    LocalAI

    mudler/LocalAI

    Run your AI models locally and generate images and audio (alternative to OpenAI and Claude).

    AI & LLM ToolsGoDocker

    AI-nativeLocalAI has 49,405 GitHub stars, uses the MIT license, ships a Docker image and had 3,438 commits in the last year; it replaces ChatGPT.

  3. Rank 3:

    LLM Harbor

    av/harbor

    Containerized LLM toolkit. Run LLM backends, APIs, frontends, and additional services via a concise CLI.

    AI & LLM ToolsPythonDocker

    AI-nativeLLM Harbor has 3,237 GitHub stars, uses the Apache-2.0 license, ships a Docker image and had 1,288 commits in the last year.

1 of 6 · 3 picks

Run models on your own hardware

Model runners that serve open-weight LLMs on your own CPU or GPU, with an OpenAI-compatible API the other tools can call.

  1. 1

    Ollama

    Get up and running with Llama 3.3, DeepSeek-R1, Phi-4, Gemma 3, and other large language models.

    AI-nativeOllama has 182,289 GitHub stars, uses the MIT license, ships a Docker image and had 1,083 commits in the last year; it replaces ChatGPT.

  2. 2

    LocalAI

    Run your AI models locally and generate images and audio (alternative to OpenAI and Claude).

    AI-nativeLocalAI has 49,405 GitHub stars, uses the MIT license, ships a Docker image and had 3,438 commits in the last year; it replaces ChatGPT.

  3. 3

    LLM Harbor

    Containerized LLM toolkit. Run LLM backends, APIs, frontends, and additional services via a concise CLI.

    AI-nativeLLM Harbor has 3,237 GitHub stars, uses the Apache-2.0 license, ships a Docker image and had 1,288 commits in the last year.

2 of 6 · 4 picks

Private chat interface

Chat front ends for your team. Point them at a local model and prompts never leave your network.

  1. 1

    LibreChat

    Enhanced ChatGPT-compatible AI chat interface supporting multiple AI providers, with multi-user auth, message search, and plugin support.

    AI-nativeLibreChat has 45,318 GitHub stars, uses the MIT license, ships a Docker image and had 2,074 commits in the last year; it replaces ChatGPT and Lovable.

  2. 2

    Open-WebUI

    User-friendly AI Interface, supports Ollama, OpenAI API.

    AI-nativeOpen-WebUI has 154,037 GitHub stars, is source-available under the Open WebUI License (BSD-3-Clause plus branding clause) (not an open-source license), ships a Docker image and had 4,746 commits in the last year; it replaces ChatGPT and Lovable.

  3. 3

    LobeHub

    Modern design AI chat framework supporting multiple AI providers, one click install MCP Marketplace and Artifacts / Thinking.

    AI-nativeLobeHub has 83,008 GitHub stars, is source-available under the LobeHub Community License (not an open-source license), ships a Docker image and had 7,488 commits in the last year; it replaces ChatGPT and Lovable.

  4. 4

    Onyx Community Edition

    Chat UI that works with any LLM. It comes loaded with advanced features like agents, web search, RAG, MCP, deep research, Connectors to 40+ knowledge sources, and more.

    AI-nativeOnyx Community Edition has 32,332 GitHub stars, uses the MIT license, ships a Docker image and had 5,771 commits in the last year.

3 of 6 · 3 picks

Questions over your documents (RAG)

Retrieval-augmented generation: index your own files and ask questions about them.

  1. 1

    RAGFlow

    Open-source RAG engine with deep document understanding.

    AI-nativeRAGFlow has 91,708 GitHub stars, uses the Apache-2.0 license, ships a Docker image and had 5,821 commits in the last year.

  2. 2

    AnythingLLM

    All-in-one desktop & Docker AI application with built-in RAG, AI agents, No-code agent builder, MCP compatibility, and more.

    AI-nativeAnythingLLM has 66,738 GitHub stars, uses the MIT license, ships a Docker image and had 887 commits in the last year; it replaces ChatGPT.

  3. 3

    Khoj

    Your AI second brain. Get answers from the web or your docs. Build custom agents, schedule automations, do deep research. Turn any online or local LLM into your personal, autonomous AI.

    AI-nativeKhoj has 37,565 GitHub stars, uses the AGPL-3.0 license, ships a Docker image and had 120 commits in the last year; it replaces ChatGPT.

4 of 6 · 4 picks

Agents and workflows

Visual builders that chain models, tools and your own APIs into agents and automations.

  1. 1

    Langflow

    Visual framework for building and deploying AI agents and RAG workflows.

    AI-nativeLangflow has 155,530 GitHub stars, uses the MIT license, ships a Docker image and had 2,486 commits in the last year.

  2. 2

    Dify.ai

    Build, test and deploy LLM applications.

    AI-nativeDify.ai has 157,921 GitHub stars, is source-available under the Dify Open Source License (modified Apache 2.0) (not an open-source license), ships a Docker image and had 6,112 commits in the last year; it replaces ChatGPT and Lovable.

  3. 3

    n8n

    Free node based Workflow Automation Tool. Easily automate tasks across different services.

    n8n has 206,738 GitHub stars, is source-available under the Sustainable Use License 1.0 (not an open-source license), ships a Docker image and had 9,029 commits in the last year; it replaces Zapier and IFTTT.

  4. 4

    Activepieces

    No-code business automation tool like Zapier or Tray. For example, you can send a Slack notification for each new Trello card.

    AI-nativeActivepieces has 24,913 GitHub stars, uses the MIT license, ships a Docker image and had 5,647 commits in the last year; it replaces Zapier and IFTTT.

5 of 6 · 4 picks

LLM observability and gateways

Trace prompts, measure cost and evaluate answers once AI features reach production.

  1. 1

    Langfuse

    LLM engineering platform for model tracing, prompt management, and application evaluation. Langfuse helps teams collaboratively debug, analyze, and iterate on their LLM applications such as chatbots…

    AI-nativeLangfuse has 35,416 GitHub stars, uses the MIT license, ships a Docker image and had 4,298 commits in the last year; it replaces Sentry.

  2. 2

    Opik

    Evaluate, test, and ship LLM applications with a suite of observability tools to calibrate language model outputs across your dev and production lifecycle.

    AI-nativeOpik has 22,399 GitHub stars, uses the Apache-2.0 license, ships a Docker image and had 3,961 commits in the last year.

  3. 3

    Agenta

    LLMOps platform for prompt management, LLM evaluation, and observability. Build, evaluate, and monitor production-grade LLM applications with collaborative prompt engineering.

    AI-nativeAgenta has 4,807 GitHub stars, uses the MIT license, ships a Docker image and had 18,955 commits in the last year.

  4. 4

    GoModel

    AI gateway written in Go with a unified OpenAI-compatible API for multiple LLM providers, USD cost tracking, budgets, usage analytics, guardrails, caching, and an admin dashboard.

    AI-nativeGoModel has 1,219 GitHub stars, uses the MIT license, ships a Docker image and had 657 commits in the last year.

6 of 6 · 4 picks

What to connect it to

  1. 1

    Metabase

    Easy way for everyone in your company to ask questions and learn from data.

    Metabase has 49,547 GitHub stars, uses the AGPL-3.0 license, ships a Docker image and had 6,378 commits in the last year; it replaces Metabase Cloud and Tableau.

  2. 2

    HyperSwitch

    Payment switch to make payments fast, reliable and affordable. Connect with multiple payment processors and route traffic effortlessly, all with a single API integration.

    HyperSwitch has 45,286 GitHub stars, uses the Apache-2.0 license, ships a Docker image and had 2,103 commits in the last year.

  3. 3

    Lago

    Metering and usage-based billing.

    Lago has 10,657 GitHub stars, uses the AGPL-3.0 license, ships a Docker image and had 174 commits in the last year.

  4. 4

    Bigcapital

    Financial accounting and inventory management software for small to medium businesses.

    Bigcapital has 3,921 GitHub stars, uses the AGPL-3.0 license, ships a Docker image and had 1,005 commits in the last year; it replaces QuickBooks and Xero.

FAQ

Questions and answers

Still curious? Email info@appsgit.com.

What open source AI tools can banks and fintech companies self-host?

For banks and fintech companies, the core of a private AI stack is Ollama for running open models locally, LibreChat for a private chat interface and RAGFlow for answering questions over your documents, all running on your own servers so customer and transaction data can stay on your infrastructure. Each pick was chosen by hand for its job and is shown with GitHub data from October 2026.

Does self-hosted AI keep my data private?

It can. When the chat interface and RAG tools are pointed at a model running on your own server (for example through Ollama or LocalAI), prompts and documents are not sent to an outside AI provider, so customer and transaction data stays on infrastructure you control. Several of these tools can also call cloud APIs, so check each connection you configure.

Does self-hosted AI make a bank compliant?

No tool does that on its own. Keeping data on your servers helps with data residency and third-party risk, but you still need model risk management, access control, logging and review under your regulator's rules. This page is not legal or compliance advice.

Is Ollama free?

Yes. Ollama is open-source software released under the MIT license, so you can self-host it at no cost. You only pay for the server it runs on, and some projects also offer an optional paid cloud plan.

Can I run this stack with Docker?

All 22 apps here publish Docker images, including Ollama, LocalAI and LLM Harbor, so most of the stack can run side by side with Docker Compose on one server. Size the server for the heaviest app and add off-site backups.

The weekly digest

New self-hosted apps, every week

Fresh releases, rising projects and one deploy guide a week. Join home labbers and engineers who self-host.

One email a week. No spam, unsubscribe anytime.