Good news, nothing change for @AiDorothy1 users, you can continue to use to Claude Code Sub, like before ππ€
Website : https://t.co/tMkOWJanii
Github (open source) : https://t.co/nkaSdS0tnL
I found @mert idea super cool, so I vibe coded the same tool for EVM wallets (and Solana ones π) using @OctavFi API.
And it's publicly available here : https://t.co/rTZKwExZRL
Grab your Octav API key and try it with any wallet π₯
DM me for free API credits to test it π
the human approval gate is the most underrated feature in agent orchestration. full automation is cool until your agent deploys a bug to prod lol. Dorothy sends you a Telegram message with context and you tap approve or reject. simple but saves you from disaster
multi-agent system where AI acts as PM, Coder and QA is not a meme anymore its just tuesday. role-based agent profiles with specialized prompts and permissions is the move. Dorothy lets you template these and chain them in automations. ship while you sleep fr
just saw someone run 10+ coding agents on the same repo with no conflicts. isolation per agent is key. git worktrees, separate branches, auto-PR. Dorothy's orchestrator handles the coordination layer so agents dont step on each other. this is the hard part nobody talks about
native macOS orchestrator > electron wrapper, every time. your battery agrees lol. Dorothy focuses on the management layer instead of wrapping terminals. kanban, automations, Telegram/Slack notifs. let the terminal be a terminal, orchestrate from above
knowledge graphs for agent orchestration is such a smart pattern. cross-agent context sharing is the hardest part of multi-agent systems. in Dorothy we use Vault docs as the context handover layer between pipeline steps but a graph model would be next level
the "Agentic OS" era is here. visual workflows, MCP integrations, kanban boards, scheduled automations... Dorothy ships all of this for Claude Code. the future of coding is managing agents not writing code
skills tell your agent WHAT to do. but who tells them WHEN to do it, in what ORDER, and what to do with the RESULTS? thats the orchestration gap. Dorothy fills it with scheduled automations, pipeline chains, and approval gates. skills + orchestration = actual productivity
hot take: agent-agnostic orchestrators sound good on paper but going deep on one runtime (Claude Code) gives you way better DX. MCP tools, native memory, tight context management... Dorothy is Claude Code-native and the integration depth is worth the tradeoff imo
everyone building parallel agent orchestrators rn but nobody talks about the management problem. 8 agents running, 3 need review, 1 stuck on CI... how do you track that? we use kanban + Telegram notifs in Dorothy. "tell the orchestrator and go for a walk" energy lol
been seeing so many people building agent orchestrators this week. the "Irfrit Loop" pattern is basically what we do in Dorothy: chain Claude Code agents (coder β reviewers β auto-fix β QA β human gate via Telegram). async agent management is the future tbh