networkevolution/appteleporter_lite repository overview

⁠What is AppTelePorter?

AppTelePorter is a container-native AI platform that gives network and DevOps teams a conversational AI agent capable of executing real infrastructure operations — with every credential encrypted in a vault that never leaves your server.

Think of it as your own ChatAI App that can actually SSH into your switches, query your APIs, and run your automation scripts — with credentials never exposed to any LLM, local or cloud.

It ships in two editions:

• Lite edition (this image) — bring your own LLM (OpenAI, Claude, or Gemini). The platform and all credentials remain on your server; conversation context is sent to your chosen cloud LLM provider for inference only. Your credentials never leave the container. Choose this when you want frontier-model quality without hosting your own GPU.
• Max edition — includes a bundled local AI model. No data of any kind ever leaves your network. Runs fully offline. GPU-enabled host recommended. → Docker Hub⁠

How it works in practice:

1. You ask a question in plain English: "What VLANs are configured on switch1?"
2. AppTelePorter figures out which tool to run and presents an approval card.
3. You approve — credentials are resolved from the vault at execution time (never serialised into prompts), the tool executes, and the answer streams back.
4. Tool execution stayed local; only the conversation context passed through your chosen LLM provider for synthesis — your credentials stayed in the vault the entire time.

Key capabilities:

⁠What is Tool Studio?

Tool Studio is AppTelePorter's built-in browser IDE for building and testing your own plugins and MCP servers — no SSH, no text editor, no container rebuild needed.

From your browser at /tool-studio you can:

• Create a new plugin — generates a skeleton tool.yaml + handler.py instantly
• Edit any file in the plugin or MCP server directory with a Monaco editor (the same editor that powers VS Code)
• Install Python packages with a single click — no container rebuild required
• Run a tool live and watch output stream in real time via the built-in console
• Hot-reload — changes take effect immediately; the model sees your updated tool on the next request
• Test MCP servers — trigger dependency installation and validate tool discovery without leaving the browser

The edit-run loop drops from "edit file → rebuild container → test" to "edit in browser → click Run."

⁠How to Install AppTelePorter Lite

Getting AppTelePorter running is as simple as a single docker run command — no configuration files, no build steps, no dependencies to install on your host machine.

⁠Step 1 — Pull the image

docker pull networkevolution/appteleporter_lite:latest

⁠Step 2 — Create a local data directory

Before starting the container, create a directory on your host machine. This is where AppTelePorter stores all its persistent data — plugins, MCP servers, the encrypted vault, chat history, and configuration.

# Linux / macOS
mkdir -p ~/appteleporter-data

# Windows (Command Prompt)
mkdir %USERPROFILE%\appteleporter-data

# Windows (PowerShell)
New-Item -ItemType Directory -Path "$env:USERPROFILE\appteleporter-data"

The Lite edition requires an external volume mount. The directory will be auto-populated with the correct folder structure on first run.

⁠Step 3 — Start the container

Ports:

• 3000 — AppTelePorter web UI (required)
• 1514/tcp and 1514/udp — Syslog receiver for Device Beacon (optional)

⁠Linux

Without syslog receiver:

docker run -d --name appteleporter \
  -v ~/appteleporter-data:/app/backend/appteleporter-data \
  -p 3000:3000 \
  --restart unless-stopped \
  networkevolution/appteleporter_lite:latest

With syslog receiver (Device Beacon):

docker run -d --name appteleporter \
  -v ~/appteleporter-data:/app/backend/appteleporter-data \
  -p 3000:3000 \
  -p 1514:1514/tcp \
  -p 1514:1514/udp \
  --restart unless-stopped \
  networkevolution/appteleporter_lite:latest

⁠macOS / Windows (Docker Desktop)

1. Open Docker Desktop and search for networkevolution/appteleporter_lite in the Search bar.
2. Click Pull to download the image.
3. Find the image under Images and click Run.
4. Expand Optional settings and fill in:
   • Container name: appteleporter
   • Ports — Host port 3000 → Container port 3000 (required)
   • Ports — Host port 1514 → Container port 1514 (optional — TCP and UDP)
   • Volumes — Host path: the directory you created in Step 2
   • Volumes — Container path: /app/backend/appteleporter-data

     ⚠️ Do not change the container path. Enter it exactly as shown.

5. Click Run.

⁠Step 4 — Open the UI and complete first-time setup

Open your browser and navigate to http://localhost:3000.

⁠4a — Set the vault master password

The first screen asks you to create the vault master password. This password encrypts everything AppTelePorter stores, including:

• All device and API credentials (SSH usernames, passwords, API keys, host addresses)
• User accounts and access tokens
• LLM provider API keys and model settings
• Application session signing secrets

⁠4b — Create the first admin user

After setting the vault password, create the first admin user — the account you will use to log in. Additional users can be added later from Settings → Users.

⁠4c — Configure your LLM provider

The AI cannot process requests until you connect it to an LLM provider. Go to:

Settings → LLM Settings

Select the model provider you want to use:

• OpenAI (ChatGPT) — enter your OpenAI API key
• Google Gemini — enter your Google AI Studio API key
• Anthropic Claude — enter your Anthropic API key

After entering the API key, click Test Connection to verify it is working.

⁠Available Plugins

These plugins are ready to drop into your AppTelePorter instance. Each one is a self-contained directory with tool.yaml and handler.py. Example scripts include detailed inline comments explaining every step of the logic.

⁠Network — Cisco

⁠Simulated / Demo Tools

⁠Utilities

⁠Available MCP Servers

MCP (Model Context Protocol) servers allow richer integrations. AppTelePorter auto-discovers and launches them; their tools appear in the same chat interface as native plugins. Tool names are prefixed: {server-name}__{tool-name}.

⁠How to Use These Tools in AppTelePorter

⁠Option A — Copy into a running container

docker cp ./plugins/cisco_switch_check \
  appteleporter:/app/backend/appteleporter-data/plugins/cisco_switch_check

curl -X GET http://localhost:3000/api/tools/reload \
  -H "Authorization: Bearer <your-api-key>"

⁠Option B — Place in your volume mount before starting

cp -r ./plugins/cisco_switch_check \
  ./appteleporter-data/plugins/cisco_switch_check

Tools are discovered automatically on next container start.

⁠Option C — Use Tool Studio (recommended)

1. Open AppTelePorter and navigate to Tool Studio.
2. Click "New Plugin" — a skeleton tool.yaml and handler.py are created instantly.
3. Paste in the plugin code from the GitHub repository⁠.
4. Click "Run" to test immediately — output streams live in the console.

⁠Installing Python dependencies

• Via Tool Studio: Click "Install Deps" — packages are installed instantly, no rebuild needed.
• Via requirements.txt: Add a requirements.txt to the plugin directory. AppTelePorter installs it automatically on container startup.

⁠Extending AppTelePorter

A plugin is two files (tool.yaml + handler.py) — drop them into your data directory and the AI picks them up immediately, no rebuild needed. An MCP server is a Python process for richer integrations (persistent connections, multi-tool servers). Full guides, templates, and examples with detailed inline comments are in the GitHub repository⁠.

⁠Why Build With AppTelePorter?

• No dev environment setup — plugins run directly inside the container; no local Python environment, no build step
• Install packages in seconds — requirements.txt or one click in Tool Studio; no rebuild, no restart
• Test live from the browser — Tool Studio's run console executes against real devices while you write the code
• The model handles parameter collection — you write the business logic; the AI collects values from the user's question
• Credentials stay in the vault — get_vault_cred() is the only credential API; no hardcoded passwords, no risk of leakage
• Very easy to implement and run — two files, a function that returns a string, exceptions caught inside; example scripts have detailed inline comments

⁠Plugin vs. MCP Server — Quick Reference

⁠What's Coming

• Network automation tools: BGP diagnostics, OSPF neighbour checks, route table queries, ACL audits
• Cloud integrations: AWS, Azure, GCP resource queries via API
• Monitoring integrations: Prometheus metrics, Grafana annotations, Zabbix host status
• Infrastructure integrations: NetBox, Nautobot, ServiceNow CMDB lookups
• Security tools: Certificate expiry checks, port scan summaries, CVE lookups
• More MCP servers: Protocol-native clients, multi-tool servers for complex workflows

Star the GitHub repository⁠ to be notified of new additions.

⁠Resources

⁠License

AppTelePorter Community Edition Lite is proprietary software provided free of charge. © 2026 NetworkEvolution. Redistribution, resale, sublicensing, or repackaging of this container image without prior written permission from NetworkEvolution is strictly prohibited.