AGENT SOFTWARE SUITE
AI coding agents for shared memory, voice control, test automation, and terminal workflows — built for Claude, Codex, and Gemini.
A suite of AI coding agents — shared memory, voice control, test harness, and terminal. Built for developers who ship with Claude, Codex, and Gemini.
THE SUITE
Agent Wispr
LIVEPress a hotkey, speak, and your words appear anywhere on screen. 100% local AI transcription powered by Whisper — no cloud, no subscription, no latency surprises.
Agent Brain
FINAL DEVSelf-hosted shared memory and semantic caching layer for AI coding agents. Connect Claude, Codex, Gemini, and Cursor to the same persistent context via MCP or REST.
Agent Harness
COMING SOONAutomated test harness for AI agents — define scenarios, run them headlessly, score outputs, and iterate on agent behaviors at scale without manual review.
Agent Terminal
COMING SOONA purpose-built terminal UI designed around AI-assisted development workflows — context-aware, agent-friendly, and fast.
AGENT WISPR IN ACTION
Watch local AI voice dictation work in any app — no cloud, no lag.
HOW WISPR WORKS
Three steps. Fully offline. No setup beyond install.
01
Press your hotkey
A single configurable hotkey activates Wispr from anywhere — browser, editor, terminal, Slack.
02
Speak naturally
Whisper runs fully on-device. No audio leaves your machine. No API key required.
03
Text appears instantly
Transcribed text is injected directly into the focused input — works in any app on your OS.
SCREENSHOTS
Stats dashboard, workspace history, and custom word dictionary.
AGENT BRAIN
Shared memory and semantic caching for your entire AI coding workflow.
40–70%
Token reduction via semantic cache
4+
AI agents share the same memory
100%
Self-hosted — zero cloud dependency
2
Integration paths: MCP + REST API
Architecture
Claude / Codex / Gemini / Cursor→MCP or REST API→Agent Brain→Qdrant · PostgreSQL · Redis
KEY FEATURES
- ▸Cross-agent memory: share context between Claude, Codex, Gemini, Cursor, and Aider across sessions.
- ▸Semantic cache: 40-70% token reduction by returning cached results for semantically similar queries.
- ▸MCP + REST API: plug into any AI coding agent via the Model Context Protocol or standard REST endpoints.
- ▸Privacy-first self-hosted: all memory stays on your machine — no cloud, no telemetry.
- ▸Local model support: run Ollama (qwen3, llama3) for embeddings without an API key.
- ▸Desktop app: Tauri 2.x desktop app with one-click Docker lifecycle control for Agent Brain.
TECH STACK
TYPESCRIPTNEXT.JSTAURIPOSTGRESQLQDRANTREDISNATSDOCKERMCPOLLAMARUSTCI/CD
Built for developers who ship with AI
Try Agent Wispr free today, or explore the Agent Brain site to follow along as it reaches launch.