Agenttalkie — AI skill for Claude Code
One voice workspace to create documents in Ambiguous AI, research with Exa, and launch Codex and Claude Code through Ori and OpenRouter.
How to install Agenttalkie
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open benikigai/agenttalkie and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Agenttalkie does
One voice workspace to create documents in Ambiguous AI, research with Exa, and launch Codex and Claude Code through Ori and OpenRouter. Watch live tool output, play generated apps, and hand the same artifact between coding agents. Built at AI Tinkerers Agents Everywhere, September 2026.
Alternatives in AI
- Part 4 — Generate AGENTS.md And AI Agent Configuration Files — I'll help you create the instruction files that will guide your AI coding assistant to build your MVP 2.1k ★
- Llmwiki — Open Source Implementation of Karpathy's LLM Wiki 1.5k ★
- Launch An AI Feature — Workflow recipe — take an AI/LLM feature from a probabilistic-aware PRD to a launch-ready model card by chaini 1.3k ★
README
AgentTalkie
One voice workspace to direct your agents and tools, follow their progress, and review the results.
Every new agent adds another chat, dashboard, or terminal to manage. AgentTalkie connects document work, research, and coding in one conversation, with actual tool output and a playable preview of what the coding agents build.
[Watch the 2:05 demo](https://agenttalkie.app/#demo) · [Open the live workspace](https://agenttalkie.app/workspace)
The recording is public. The live workspace requires a demo access code and configured providers.
What we built
The recorded workflow uses four integrations in the same conversation:
- Ambiguous AI: draft a Personal EA job description, approve the save, and open the resulting document in Ambiguous.
- Exa: retrieve public research and inspect the returned source references.
- Codex through Ori: create a playable tic-tac-toe game in a fresh workspace, with native CLI output visible in the dashboard.
- Claude Code through Ori: edit the exact HTML file Codex produced, changing the game colors while retaining its behavior.
The generated app runs inside AgentTalkie. Each successful version has a content hash and parent artifact reference. The previous working preview stays available while an edit runs or if it fails.
The app also supports direct Ambiguous task creation and updates with approval and readback, conversation history, new conversations, question corrections, and explicit pending, failed, and unknown outcomes. A proposed action is shown separately from a completed tool call.
How it works
flowchart TB
subgraph Browser["Browser"]
UI["Next.js / React workspace
Voice dock, transcript, activity, live output"]
Preview["Playable HTML preview
Sandboxed iframe"]
end
Voice["OpenAI GPT-Live
WebRTC voice session"]
UI <-->|"Audio and conversation events"| Voice
subgraph Cloud["Vercel application backend"]
API["Authenticated API
Voice broke
Related Skills
Validate Voice
Run the Pass-C checks on generated output (facts, AI tells, repetition)
Bench Watch
Launch or attach to a Plumbline benchmark slice, poll it to completion, and emit the canonical anti-Goodhart p
Brainstorm Loop
In-session loop — watch a folder for AI brainstorm session documents and advance each per the shared protocol
Legacy To Wavemaker Claude Plugin
Claude Code plugin to migrate legacy web apps (JSP/servlet & more) to WaveMaker AI and hand-author WaveMaker f
Knowledge Distiller
AI agent skill that distills documents into temporally-aware structured knowledge graphs. Dual output: .knowle
Review Video
Review a rendered video with AI vision. Uses TwelveLabs to watch the output, analyze pacing/composition/audio
Related Agents
Flutter Android Deployment
Use this agent when deploying Flutter apps to Google Play Store. Specializes in Android app signing, Google Pl
Personal Ingest
Ingests personal/private material — meeting notes (Granola, pasted, hand-summarized), voice memos (.opus / .m4
Envoy Runner
Proactively use when Claude Code is stuck, wants a second implementation or diagnosis pass, needs a deeper roo