esaradev

esaradev / icarus-plugin

Public

Self-memory and replacement models for Hermes agents. Remember your work. Train your replacement.

40
1
100% credibility
Found Apr 01, 2026 at 40 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Icarus is a plugin for Hermes AI agents that provides shared persistent memory through notes, automatic decision capture, and tools to fine-tune and switch to custom replacement models based on the agents' own history.

How It Works

1
🔍 Discover smarter AI helpers

You hear about Hermes, a chat tool for AI agents that can remember their work and get better over time with a special add-on called Icarus.

2
📥 Add Icarus to Hermes

Copy the Icarus files into your Hermes folder so your agents can start sharing memories right away.

3
📁 Pick a notes folder

Choose a simple folder on your computer where all your agents' notes and memories will be saved and shared.

4
💭 Chat and watch memories form

Start talking to your agent in Hermes – it automatically saves key decisions, reviews, and fixes as neat notes you can see grow.

5
🔄 Review and hand off work

Your agents pass tasks to each other, linking notes so everyone sees the full story of what was built, reviewed, and improved.

6
🚀 Train your own smarter model

Once you have enough great notes, turn them into a custom AI brain that's trained on your own work and ready to use.

✅ Agents remember and improve forever

Now your team of AI agents shares perfect memory, works together seamlessly, and runs on cheaper, personalized smarts – everything flows better.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 40 to 40 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is icarus-plugin?

Icarus-plugin is a Python plugin for Hermes agents that delivers shared self-memory across sessions and profiles via markdown notes in a configurable directory. Agents recall past work with ranked search, write linked entries for handoffs like reviews and fixes, and hooks automatically capture decisions, outcomes, and session summaries. It solves agents forgetting context between runs by turning notes into training data for fine-tuning replacement models on Together AI, complete with eval and switch tools.

Why is it gaining traction?

Unlike basic per-session memory in Hermes, icarus-plugin enables cross-agent workflows where a coder hands off to a reviewer via pending tasks and linked notes, all viewable in Obsidian. The killer hook is the end-to-end replacement model pipeline: export high-precision pairs from your history, train a cheaper model, eval it against your fabric prompts, and swap if it passes—agents literally train their successors. With tools like fabric_brief for daily ops and telemetry for recall quality, it feels like a production memory layer for agent teams.

Who should use this?

Hermes users building multi-profile agent teams, like a "coder" finishing rate limiters and a "reviewer" spotting races, needing shared recall without orchestration bloat. Python devs running persistent Hermes chats who want to distill their agent's work into fine-tuned models for cost savings. Ideal for icarus server plugins setups where agents handle code reviews, bug fixes, or customer tickets across instances.

Verdict

Grab it if you're deep in Hermes and crave persistent memory plus self-improving models—docs are thorough, quickstart works, and tests cover handoffs and exports. At 40 stars and 1.0% credibility, it's early-stage raw; expect tweaks as multi-agent patterns mature, but the concept nails agent evolution.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.