RitikPatill

Waypoint Agents — Design skill for Claude Code

Design community

Durable, event-sourced multi-agent orchestrator with live timeline UI.

How to install Waypoint Agents

This entry records only its repository, not the path inside it, so there is no exact command to give. Open RitikPatill/waypoint-agents and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Waypoint Agents does

Durable, event-sourced multi-agent orchestrator with live timeline UI. Define agent workflows, run them, kill the process, restart—Waypoint replays its event log to resume from the last checkpoint without re-running completed steps or re-charging for LLM calls.

Alternatives in Design

README

Waypoint — Durable Multi-Agent Workflows with a Visual Timeline

**Video walkthrough:** https://youtu.be/RJcQt1JwUP0 **60-second overview:** https://youtu.be/nEWEkPkN_co

Event-sourced multi-agent orchestrator: kill it mid-run, restart it, watch it resume from the last checkpoint in a live timeline UI.

![demo](docs/demo.gif)

What it is

Waypoint is a reference implementation of durable agent orchestration. You define a workflow as a directed graph of cooperating agents; the engine executes it while appending every LLM call, tool execution, and handoff to a SQLite event log before treating each step as done. Kill the process at any point, restart it, and Waypoint replays the log to resume from the exact next pending step — no repeated API charges, no double-executed tools.

The project ships two example workflows (a two-agent research team and a three-agent code reviewer), a FastAPI + SSE backend that streams events to the browser as they land, and a lightweight timeline UI with per-agent swimlanes, click-to-expand event cards, and a "Kill Worker" button for the crash-resume demo.

Quickstart

git clone https://github.com/RitikPatill/waypoint-agents.git
cd waypoint-agents
uv sync
cp .env.example .env
# edit .env — set ANTHROPIC_API_KEY
uv run waypoint serve

Open `http://localhost:8000`. To run without Docker or uv, see the Docker path below:

cp .env.example .env
docker compose up

Tests pass without an API key:

uv run pytest

Usage

Open `http://localhost:8000`. The run form offers two workflows from a dropdown.

**Crash-resume demo (research_team)**

Select `research_team`, enter a prompt such as `State of MCP servers in 2026`, and click Run. The timeline shows the Planner agent reasoning and handing off to the Researcher, which begins streaming `fetch_url`