AI Agent

Intelligent assistance,
built right in

A VS Code-style chat panel backed by any major LLM provider — with RAG, function-calling skills, streaming responses, and long memory. All inside MedICS.

Multi-Provider Support

Connect to any
LLM provider

Bring your own API key or use a local model via Ollama. MedICS normalises all providers to a single interface.

Supported LLM Providers

OpenAI
Gemini
Anthropic
Grok
GitHub Copilot
Ollama
Custom Endpoints
Streaming responses — token-by-token rendering at 60 fps
RAG — BM25 retrieval over curated docs & extension metadata
Conversation history — sessions saved & restored across restarts
Long memory — persistent user facts in MEMORY.md
AutoGen multi-agent — orchestrate complex workflows with autogen-agentchat

Built-in Skills

run_in_console Execute code in the visible Jupyter console
get_workspace_info List workspace variable names and types
get_current_state App state: folder, file, theme, extensions
workspace Read/write workspace variables by name
long_memory Persist user facts to MEMORY.md
create_skill Scaffold and install a new agent skill
agent_memory In-session temporary key-value store
AutoGen Integration

Multi-agent workflows
for complex analysis

MedICS integrates autogen-agentchat ≥ 0.7.5, enabling multi-agent orchestration, tool-use loops, and autonomous analysis pipelines.

🔄

Tool-Use Loops

Agents autonomously call skills — running Python code, reading workspace variables, and retrieving documents — until the task is complete.

👥

Multi-Agent Teams

Orchestrate specialist agents: a planner, a coder, and a reviewer collaborating on complex imaging analysis tasks.

📚

RAG-Augmented Context

BM25 retrieval over curated MedICS docs, extension metadata, and your own knowledge files injected automatically into context.

🧠

Persistent Memory

User facts, preferences, and project notes survive session restarts via MEMORY.md — the agent remembers who you are.

Extensible Skills

Build your own
agent skills

Skills are Python callables exposed to the agent via function-calling. Scaffold a new skill with one command.

bash 
# Scaffold a new skill
medics --create-skill my_analysis_skill
python 
from medics.skills import skill

@skill(name="my_analysis", description="Run my custom analysis on the active image.")
def my_analysis(app_context, param: str) -> str:
    img = app_context.workspace["image"]
    result = do_something(img, param)
    return f"Analysis complete: {result}"

Start using AI in your imaging workflow

Install MedICS and connect your preferred LLM provider in minutes.

Install MedICS → Explore Platform