Capture tasks and reminders through a phone call or Telegram voice note instead of a to-do app.
Get a task automatically broken down into one small next physical step to reduce start-up friction.
Have specialist agents handle browsing, bookkeeping, and multi-step planning behind one voice interface.
Watch the live agent network, memory, and productivity trends through the included dashboard.
| jamesbrownvc/buddy | 1038lab/comfyui-agnes-ai | andyuneducated/resolve-ai | |
|---|---|---|---|
| Stars | 18 | 18 | 18 |
| Language | Python | Python | Python |
| Setup difficulty | hard | easy | hard |
| Complexity | 5/5 | 2/5 | 4/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Runs a full local mesh of agents, a hub, and a containerized browser, requiring several services set up on local hardware.
Buddy is a voice first productivity system built for people with ADHD, framed as an external executive function rather than a normal to-do app. The README explains that ADHD is not a knowledge problem: people usually know what they should do, but struggle with actually starting tasks, holding a plan in working memory, judging how much time has passed, and following through without shame when something slips. Ordinary productivity tools assume the exact abilities ADHD affects, which is why they tend to get abandoned. Buddy's approach is to live where the user already is, such as a phone call or a Telegram voice note, capture whatever is said in loose half sentences, hold the plan on the user's behalf, break tasks down into one small physical next step, nudge at the moment action is actually needed, and treat a missed task as information rather than a failure. Under the hood, Buddy is a network of small, single purpose agents connected to one central hub, running entirely on local hardware. Every agent exposes the same simple contract, accepting a message and returning a reply, which lets agents call each other freely. A voice agent built on ElevenLabs, playing the role of an evidence based ADHD companion, can hand off parts of a conversation to any of the other specialist agents mid call. Those specialists include a bookkeeper agent that manages the user's memory, a browser agent that drives a real browser to complete web tasks, and an orchestrator that routes larger multi step requests to planning and building agents. Each specialist runs as its own independent autonomous agent with its own persona and tools, rather than following hard coded logic written in Python, and agents decide for themselves when to delegate work to a peer over these tool based connections. A dashboard shows the live network of agents, weekly productivity trends, and lets a user click into any agent to see its documentation, memory, or live browser session. The project won the Hermes Buildathon in 2026.
A voice first network of autonomous agents built to act as an external executive function for people with ADHD, remembering tasks and nudging follow-through.
Mainly Python. The stack also includes Python, FastAPI, ElevenLabs.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly general.
This repo across BitVibe Labs
double-check against the repo, no cap.