Workshop | Raindrop — Debug your AI agent locally

/workshop
JUST LAUNCHEDGitHubDOCS
Compatible withTypeScript·Python·Rust·Go·Vercel AI SDK·OpenAI SDK·Anthropic SDK·LangChain·LlamaIndex·CrewAI·Mastra·Claude Code CLI·Pi·OpenCode·Cursor·
TypeScript·Python·Rust·Go·Vercel AI SDK·OpenAI SDK·Anthropic SDK·LangChain·LlamaIndex·CrewAI·Mastra·Claude Code CLI·Pi·OpenCode·Cursor·
TypeScript·Python·Rust·Go·Vercel AI SDK·OpenAI SDK·Anthropic SDK·LangChain·LlamaIndex·CrewAI·Mastra·Claude Code CLI·Pi·OpenCode·Cursor·
Workshop by RaindropGive Claude CodeCodexDevinCursorOpenCodeClaude Code the power to write and run agent evals.
Free. Local. Open-source.
the local debugger your agent is missing.Star on GitHub2.4kJUST LAUNCHEDRead the Docs
macOS, Linux, and probably Windows.
watch your agent think.
◀▶↻
🔒localhost:5899
connectedclear
Search runs...
vet_assistant (aa7eb)
just now
vet_assistant (74d67)
3m ago
claude_code_session (fd183)
4m ago
claude_code_session (63003)
7m ago
vet_assistant (5ace0)
12m ago
vet_assistant (e3040)
18m ago
claude_code_session (a6a0c)
22m ago
claude-sonnet-41 tool|DURATION ...|850 in / 210 out|$0.032TRACEaa7ebba8
OverviewSpan TreeComms
▸ System 0.2k
You are a veterinary triage assistant. Assess symptom severity, check the symptom database, and recommend whether the pet needs immediate, same-day, or routine care. Always ask clarifying questions.
🐾 PawsAI — Veterinary Assistant
🐾PawsAIbeta
▌
Describe your pet's symptoms...
claude code
user@dev ~/pawsai % cd pawsai
user@dev ~/pawsai % claude
Claude Code v2.1.3
Sonnet 4 · Claude Max
/Users/dev/pawsai
> █
🐾 PawsAI
🐾PawsAIbeta
▌
Describe your pet's symptoms...
workshop
connectedclear
Search runs...
vet_assistant (aa7eb)
just now
vet_assistant (74d67)
3m ago
claude_code_session (fd183)
4m ago
claude_code_session (63003)
7m ago
vet_assistant (5ace0)
12m ago
vet_assistant (e3040)
18m ago
claude_code_session (a6a0c)
22m ago
claude-sonnet-41 tool|DURATION ...|850 in / 210 out|$0.032TRACEaa7ebba8
OverviewSpan TreeComms
▸ System 0.2k
You are a veterinary triage assistant. Assess symptom severity, check the symptom database, and recommend whether the pet needs immediate, same-day, or routine care. Always ask clarifying questions.
claude code
user@dev ~/pawsai % cd pawsai
user@dev ~/pawsai % claude
Claude Code v2.1.3
Sonnet 4 · Claude Max
/Users/dev/pawsai
> █
🔒 localhost:5899
connectedclear
Search runs...
vet_assistant (aa7eb)
just now
vet_assistant (74d67)
3m ago
claude_code_session (fd183)
4m ago
claude_code_session (63003)
7m ago
vet_assistant (5ace0)
12m ago
vet_assistant (e3040)
18m ago
claude_code_session (a6a0c)
22m ago
claude-sonnet-41 tool|DURATION ...|850 in / 210 out|$0.032TRACEaa7ebba8
OverviewSpan TreeComms
▸ System 0.2k
You are a veterinary triage assistant. Assess symptom severity, check the symptom database, and recommend whether the pet needs immediate, same-day, or routine care. Always ask clarifying questions.
claude code
user@dev ~/pawsai % cd pawsai
user@dev ~/pawsai % claude
Claude Code v2.1.3
Sonnet 4 · Claude Max
/Users/dev/pawsai
> █
🔒 localhost:5899
connected
vet_assistant (aa7eb)
just now
EVALStest_symptom_followup.ts3/3 passed
OverviewSpan TreeEvals
✓Agent asks about weight bearingpassed
✓Agent asks about swellingpassed
✓Agent asks about exercise historypassed
✓Agent doesn't jump to diagnosispassed
✓Agent recommends vet visitpassed
5 assertions·2.1s total·all passing
claude code
> the vet agent isn't asking follow-up questions
● Reading trace aa7eb from Workshop...
Agent skipped follow-ups. Writing eval.
→ Writing eval: test_symptom_followup.ts
→ Fixing: src/prompts/vet.ts:23
→ Re-running agent...
✓ eval passes after fix
✓ re-running agent...
✓ agent now asks 3 follow-up questions
01
Live streamed tracesEvery token, every tool call, every decision — streaming into Workshop the moment it happens. No polling, no refreshing.
02
Integrates with your coding agentClaude Code reads your traces, writes evals, and fixes what's broken.
03
Self-healing eval loopClaude writes the eval, runs your agent, sees the failure, fixes the code, and re-runs — until every assertion passes.
01
Live streamed tracesEvery token, every tool call, every decision — streaming into Workshop the moment it happens. No polling, no refreshing.
localhost:5899
Trajectory
ai.generateText
1.1s
check_symptoms
280ms
get_breed_risks
140ms
02
Integrates with your coding agentClaude Code reads your traces, writes evals, and fixes what's broken.
claude code
● Reading trace aa7eb...
→ Writing eval: test_symptom_followup.ts
→ Fixing: src/prompts/vet.ts:23
✓ eval passes — agent now asks follow-ups
03
Self-healing eval loopClaude writes the eval, runs your agent, sees the failure, fixes the code, and re-runs — until every assertion passes.
localhost:5899 — evals
✓Agent asks about weight bearing
✓Agent asks about swelling
✓Agent asks about exercise history
3/3 passed
342 left
145 left
Limited Edition
get the exclusive launch drip.Install Workshop and run the magic command. Free merch shipped worldwide while supplies last.
1
2
Compatible with everything
Vercel AI SDK
OpenAI Agents
Anthropic SDK
LangChain
LangGraph
CrewAI
Pydantic AI
Mastra
DSPy
DSPy
Google ADK
Strands
Strands
agno
Agno
da
Deep Agents
TypeScript
Python
Go
Rust
AWS Bedrock
Azure OpenAI
Vertex AI
Claude Code CLI
MCP
Temporal
{ }
HTTP API
oc
OpenCode
Get started
Free. Local. Open-source. One-line install.Star on GitHub|Documentation|@raindrop_ai
raindrop.ai