kevinelliott/agentpipe — A CLI/TUI app that orchestrates multi-agent conversations by enabling different AI CLI tools (Claude…

Description and use

**Brief Summary**
*agentpipe* is a Go‑based CLI/TUI that lets multiple AI agents (Claude Code, Gemini, Qwen, etc.) converse in shared “rooms,” turning ad‑hoc prompts into repeatable, orchestrated workflows. By wiring together disparate AI‑CLI tools, it provides a lightweight orchestration layer for multi‑agent coordination, tool‑use pipelines, and persistent agent memory.

**Value**

• **Workflow unification** – Isolated AI tools become collaborative agents that can pass messages, results, and context to one another without writing custom glue code.
• **Repeatability & standardisation** – Defined rooms act as reusable “conversation scripts,” making it easy to version, test, and share complex agent pipelines across teams.
• **Rapid prototyping** – The TUI gives immediate visual feedback on message flow, helping developers iterate on multi‑agent designs faster than building bespoke APIs.

**Practical Adoption Path**

1. **Pilot the CLI** – Install the binary (go install github.com/kevinelliott/agentpipe@latest) and spin up a test room with two agents (e.g., Claude and Gemini) to verify basic messaging.
2. **Integrate existing AI‑CLI tools** – Add the required agent executables to the agentpipe config (JSON/YAML). Because the tool works through standard input/output, most AI‑CLI wrappers can be dropped in with minimal adaptation.
3. **Define reusable pipelines** – Encode common sequences (e.g., “code generation → static analysis → refactor”) as room scripts or shell wrappers, committing them to a repository for team consumption.
4. **Persist memory** – Enable the built‑in memory store or hook an external KV store (Redis, SQLite) to keep context across sessions, then promote the setup to a staging environment.
5. **Scale to production** – Containerise the agentpipe process, expose its API/CLI via a sidecar or internal service, and integrate with CI/CD pipelines for automated execution of agent workflows.

**Production Readiness**

• **Activity & community** – 136 ★, 23 forks, recent commits (last updated 2026‑06‑27), and a Go codebase make the project easy to audit and extend.
• **Ecosystem fit** – The tool exposes clear API/SDK/CLI signals, works with any language‑agnostic AI CLI, and already supports multiple major models, reducing integration friction.
• **Stability** – The core orchestration logic is small and self‑contained; the primary risk areas are licensing verification, security hardening of the underlying AI binaries, and ensuring maintainers remain active.
• **Readiness level** – High for an OSS candidate; it is mature enough for a serious pilot in internal tooling or a production‑adjacent service, provided the final security and license review is completed.

Problem

kevinelliott/agentpipe helps turn isolated prompts and tools into repeatable agent workflows.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• coordinate multi-agent workflows
• add tool-use pipelines
• standardize agent memory

Evidence

A CLI/TUI app that orchestrates multi-agent conversations by enabling different AI CLI tools (Claude Code, Gemini, Qwen, etc.) to communicate in shared rooms.

Source