A self-hosted AI agent platform that turns a kanban board into an autonomous development team. Describe tasks, the manager picks them up, writes code, and delivers results for your review — all on your own hardware, with your own models, and zero data leaving your server. Free for personal use.
- Kanban-driven agents — create tasks, the manager picks them up and works autonomously
- Human-in-the-loop — every result lands in a review column before it's accepted
- Assistant / brain-dump chat — describe an idea in plain language, the system creates structured tasks
- Scheduled triggers — cron-based agent execution without manual intervention
- VS Code integration — MCP server exposes tasks and review directly in your editor
- Bring your own model — connect any OpenAI-compatible endpoint (LocalAi, LiteLLM, …), OpenAI / Azure, or the Anthropic API directly
- Claude Code CLI bridge — on a personal install, route agents through your local
claudeCLI for free with your existing Claude subscription (personal machine only — see caveat below) - 100 % self-hosted — your data never leaves your machine
- 0 % telemetry — no tracking, no analytics, no data collection ... at least not in our code 😄
curl -fsSL https://raw.githubusercontent.com/data-ps-gmbh/Agent-Elno/main/get-agent-elno.sh | sudo bashThe interactive installer asks for your LLM provider, model name, and ports — then starts everything. Same command to update. Default login: admin / password.
Already know your config? Skip the wizard with the non-interactive install template — a pre-fillable script you can also drop into cron for nightly auto-updates.
curl -fsSL https://raw.githubusercontent.com/data-ps-gmbh/Agent-Elno/main/remove-agent-elno.sh | sudo bashRemoves all services, binaries, config, and data. Asks for confirmation before deleting.
You can feed work into the system from three places — pick whichever fits your moment:
- Web kanban board — open the board, add a task with a title and description.
- Assistant chat — brain-dump an idea in plain language; the personal agent extracts intent and creates a structured task for you.
- VS Code / Copilot via MCP — describe the work in your editor; Copilot hands it into the system.
Once tasks are on the board, the autonomous loop takes over:
- The manager runs on a cron heartbeat — surveys the board, assigns each eligible task to the right worker agent (developer, senior-developer, reviewer, architect, …).
- The worker wakes up immediately, checks out the project's feature branch, writes code, runs tests, commits.
- When the worker is done, the manager evaluates the result on its next heartbeat — moves trivial work to Done, escalates substantive work to Review.
- You review — approve via comment, or request changes in a comment and remove the
reviewtag; the manager picks it up again on the next heartbeat.
The manager picks up tasks based on column and tags:
- Processed: Any column except Backlog and Done, without the tags blocked or review
- Skipped: Tasks in Backlog, Done, or tagged blocked / review
Move a task to Ready (or any active column) to let the manager pick it up. When the manager finishes, it tags or moves the task to review or done — you check the result and either approve or comment.
Away from VS Code? Use the chat to tell your personal agent things on the go — via the web UI or the mobile app (currently in closed beta).
The model is everything. Agent-Elno is an orchestration layer — it sets the stage, but the model does the actual thinking. A capable model will surprise you. A weak one will burn through your tokens producing slop. Pick the model first, configure the platform second.
Running on your own dev machine? You can wire up the Claude Code CLI as an LLM backend — the platform spawns the local
claudebinary and piggybacks on your existing Claude subscription. No API key, no per-token bill on top of what you already pay. Caveat: per Anthropic's TOS the subscription-tied CLI login is only legal on your own developer machine — not on a shared or production server. For server installs, use the Anthropic API server type with an API key. See docs/claude-setup.md.
This is the live setup at DATA-PS — it changes as new models ship, so treat it as a snapshot, not a fixed recipe.
| Use case | Model | Why |
|---|---|---|
| Coding (default developer) | Claude Sonnet 4.6 | Best size / quality / cost balance for code; reliable tool calling |
| Heavy lifting (architect, reviewer, multi-file refactors) | Claude Opus 4.7 | Strongest reasoning, huge context; catches what coder models miss |
| Manager + personal assistant (orchestration & chat) | Qwen3.6-27B (local) | Reliable tool calling, surprisingly good at code/architecture for triage and "simple-change" reviews, great secretary persona, free, private |
| Specific cloud coder tasks | gpt-5.1 (codex variants) | Where we route the work that doesn't go to Claude — large context, fast |
| Embedding | nomic-embed-text-v1.5 (local) | 768-dim, fast, good semantic search quality |
We bind these directly or front them with LiteLLM as a unified proxy.
These are the durable takeaways — model names will rot, but the patterns hold.
- Below ~30B params is too small for orchestration. Anything in the 8B–14B range loops on tool calls under load (Qwen3-8B, Phi-4-Mini variants, Hermes-4-14B, GPT-4.1-Nano). 27–32B is the floor for a reliable manager / personal agent.
- Local coder models choke on real codebases. We ran our production repos (50k–100k LOC C#, 20k–40k LOC Razor) past every flavour of Qwen-Coder, NextCoder, DeepSeek-V3.2 — all of them either ignored conventions, hallucinated paths, or "reviewed instead of coded." The working pattern is local manager + cloud (or Claude CLI) coder, not local everything.
- Reasoning models for review pay for themselves. Whatever the current frontier reasoning model is (today: Opus 4.7 / o-series), it catches what coder-tier models miss. Not a place to economise.
- Tool-call quality matters more than benchmark scores. We've watched models with great HumanEval numbers loop endlessly because they couldn't pick the right tool. Test in-platform, not on leaderboards.
- Local (LocalAI / Ollama): free, private, no rate limits — but needs a GPU (a 27–32B model wants ~24 GB VRAM at Q4).
- Cloud API (OpenAI / Azure / Anthropic API): faster, smarter coding models — costs money and data leaves your server, but it's the only legal Claude path on a shared host.
- Claude CLI (personal machine only): free if you already have a Claude subscription; see the legal caveat above.
- Hybrid via LiteLLM: route orchestration locally, coding to cloud or Claude — what we do day-to-day.
| Quick Start | Installation and first steps |
| Configuration | Environment files, config modes, all options |
| Architecture | Service architecture and data flow |
| Manager Process | How the autonomous loop works |
| Agents & Skills | Agent definitions and prompt templates |
| Chat & Memory | Personal agent, sessions, semantic memory |
| Scheduler | Cron-based triggers |
| Integrations | LiteLLM, Ollama, OpenAI, Anthropic, Claude CLI, nginx, Traefik |
| Claude Code CLI Setup | Personal-machine route to Claude (subscription, no API key) |
| Troubleshooting | Logs, action log, common issues |
| Changelog | Release history |
- Debian 12+ or Ubuntu 22.04+ (x64)
- An OpenAI-compatible LLM endpoint (Ollama, LiteLLM, OpenAI, …)
- 2 GB RAM, 4 GB disk minimum
No Docker, no .NET SDK, no runtime installation required.
We built Agent-Elno to fit our own workflow — but we're actively developing it further. If you have suggestions, feature requests, or run into problems, we'd love to hear from you.
- Feature ideas — open an issue; if it fits our roadmap, we'll try to integrate it
- Bug reports — please include a brief description of the problem and steps to reproduce; we'll investigate anything we can reproduce ourselves
- Questions — we try to answer as fast as we can
We can't promise everything, but we read every issue and do our best to help.
- GitHub Issues — bug reports, feature requests, questions
- data-ps.de — company website
- info@data-ps.de — commercial inquiries
Free for personal, non-commercial use under the PolyForm Strict License 1.0.0.
Commercial use requires a separate license — contact info@data-ps.de.
© 2014–2026 DATA-PS GmbH. All rights reserved.