The open-source AI
development environment
for your terminal.
Build. Run. Review.
From anywhere.
$pipx install andromity
Open Source · MIT · Local Models · MCP · Multi-Agent · Remote Access
The open-source AI
> Build a REST API for this project
Why Andromity?
Terminal-native
Everything you need, inside one environment. Resides in your terminal panel directly alongside your favorite text editors.
Open source
MIT licensed. Inspect the code, customize the orchestration, add custom profiles, or contribute upstream.
Bring your own model
Use your own API keys for cloud endpoints (Anthropic, OpenAI, Gemini) or run 100% offline local models (Ollama).
Work from anywhere
Your local rig remains your execution host, while you securely issue prompts and review changes from any browser.
One terminal. Your entire workflow.
Andromity acts as an orchestrator sitting between your codebase, external MCP tools, and LLM reasoning models.
Read
→
Plan
→
Approve
→
Build
→
Test
→
Review
This workflow represents our core safety design: in **SAFE mode**, file edits, creations, and command execution require your explicit verification. The agent never acts without consent.
Your codebase
↓
Andromity
↓
Agent
MCP
Models
↓↓↓
Tools
Code
Tests
Core Capability Showcase
Multi-Agent Collaboration
Runs specialized agent profiles sequentially. The **Planner** writes plans, the **Coder** performs line-edits, the **Tester** runs tests, and the **Reviewer** audits the diffs.
MCP Smart Lazy-Loading
Connects to any Database, GitHub API, or search tool. Fully indexes 50+ tools but loads full JSON parameters only on-demand, reducing token costs by **85%**.
Background & Cron Tasks
Schedule Andromity to run test suites, monitor container health, or inspect log endpoints periodically, alerting you only if exceptions occur.
Bring your own intelligence.
Compatible with LiteLLM, allowing instant connection to any major provider.
ANTHROPIC
OPENAI
GEMINI
OPENROUTER
OLLAMA
GROQ
NVIDIA NIM
Your code. Your models. Your keys. Your machine.
You are always in control.
Can you trust an AI agent running commands on your machine? With Andromity, you decide the level of access you grant.
SAFE MODE
Maximum verification. Full review controls.
The default installation mode. Andromity checks in before modifying files or running shell actions. Safe, predictable, and auditing-friendly.
File Modifications:Requires confirmation
Shell Commands:Requires confirmation
Destructive Actions:Requires confirmation
> refactor database.py
Planner analyzed 4 models.
⚡ ⚡ WRITE_FILE REQUEST: src/database.py
Do you want to write these changes? [Approve] / [Reject]
Your terminal. Everywhere.
Run Andromity securely on your developer computer, and monitor or execute tasks from your phone or secondary devices.
YOUR PCRunning locally
WEB BRIDGEE2E Encrypted relay
andromity-remote
Session: local-rig-921
> /status
✓ Agent: Idle
✓ Tests: 44 passed
> run tests and build
Executing pytest...
✓ 44 passed in 0.8s
✓ npm run build passed
_
Type command...
From Idea to Code
A straightforward progression from instruction to implementation.
01
Point Andromity
Launch the CLI inside your workspace project folder.
02
Choose model
Pick local models or export provider API keys.
03
Describe task
Type what you want to build in standard English.
04
Agent explores
Andromity scans files, symbols, and references.
05
Create plan
The Planner outputs files to change or create.
06
Review changes
Audit unified diffs line-by-line in the TUI.
07
Approve
Authorize the agent to commit modifications.
08
Run tests
Verify execution. Auto-fix bugs if tests fail.
09
Ship
Merge changes and commit to Git. Done.
Designed for Developers
Main TUI panel
[Plan] Implement API routes
[Diff] src/api.py +12 lines
[Chat] Enter command...
[Status] Running pytest...
[Diff] src/api.py +12 lines
[Chat] Enter command...
[Status] Running pytest...
Unified Terminal View
Manage conversations, plans, file diffs, and settings side-by-side inside your TUI.
Model picker (Ctrl+L)
> anthropic/claude-sonnet
openai/gpt-4o
google/gemini-1.5
ollama/qwen-14b
openai/gpt-4o
google/gemini-1.5
ollama/qwen-14b
Instant Model Switcher
Switch between cloud providers or local execution models on the fly.
MCP Tools Panel
⚡ DB: sqlite-db/query
⚡ Git: github/create_pr
⚡ Search: web_search/query
✓ Loaded 18 tools dynamically
⚡ Git: github/create_pr
⚡ Search: web_search/query
✓ Loaded 18 tools dynamically
MCP Inspector
Track connection states and verify all external capabilities active in your workspace.
See how Andromity compares.
Understand how Andromity differs from Claude Code, Aider, Gemini CLI, and Cursor.
Built in the open.
MIT licensed. Community driven. No vendor lock-in.
★ 16GitHub stars
⑂ 4Contributors
v0.1.0b2Latest Beta release
Explore User Projects
See real-world API microservices, adapters, and deployment scripts built using Andromity.
Latest from Andromity
August 15, 2026
Building a Multi-Agent Terminal Environment
How we designed a terminal-native multi-agent architecture where builder, planner, coder, and reviewer profiles collaborate autonomously.
Read Article →August 12, 2026How MCP Lazy-Loading Works
A deep dive into indexing 50+ connected MCP server schemas compactly and fully loading tool definitions only when required.
Read Article →August 08, 2026Running Local LLMs with Andromity
How to set up and configure local models like Llama 3, Qwen 2.5, and Ollama in Andromity for complete data privacy.
Read Article →