Quickstart

Create an agent, run it, and watch the real-time event stream — in 5 minutes.

Verify your stack first: lantern doctor
Before running the steps below, confirm the stack is healthy:

# Build the CLI once, then run the readiness check:
( cd packages/cli && go install ./cmd/lantern )
lantern doctor

All checks must show ✓ before continuing. Need the stack running first? Follow the Installation guide (2 min).

Get a token

The dev stack seeds an admin account. Export a JWT:

export LANTERN_TOKEN=$(curl -s -X POST http://localhost:8080/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email":"admin@lantern.dev","password":"lantern"}' \
  | grep -o '"token":"[^"]*"' | cut -d'"' -f4)

echo $LANTERN_TOKEN   # should print a JWT

Add an LLM provider

Lantern routes all model calls through its model router — add at least one key:

curl -s -X POST http://localhost:8080/v1/settings/llm-providers \
  -H "Authorization: Bearer $LANTERN_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"provider":"anthropic","api_key":"sk-ant-YOUR_KEY_HERE"}'

# Verify
curl -s -X POST http://localhost:8080/v1/settings/llm-providers/anthropic/test \
  -H "Authorization: Bearer $LANTERN_TOKEN"
# → {"ok":true}

OpenAI works too. Use "provider":"openai" with your OpenAI key. Any configured provider is available.

Create an agent

The system prompt field is systemPrompt — not instructions.

curl -s -X POST http://localhost:8080/v1/agents \
  -H "Authorization: Bearer $LANTERN_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-first-agent",
    "systemPrompt": "Answer questions clearly and concisely."
  }' | jq .

Model routing is automatic. The router picks the best configured provider. Use "reasoning-large" or "reasoning-small" capability tiers in run input to influence routing. Never hard-code vendor model names.

Run the agent

The run body uses agentName (camelCase):

export RUN_ID=$(curl -s -X POST http://localhost:8080/v1/runs \
  -H "Authorization: Bearer $LANTERN_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "agentName": "my-first-agent",
    "input": {"message": "What are the three laws of thermodynamics?"}
  }' | jq -r '.id')

echo "Run ID: $RUN_ID"

Stream the events

Every run emits live events over SSE — tool calls, model responses, completions:

curl -N http://localhost:8080/v1/runs/$RUN_ID/events \
  -H "Authorization: Bearer $LANTERN_TOKEN" \
  -H "Accept: text/event-stream"

Example output (event kinds emitted by the plain-LLM path):

event: step_started
data: {"seq":1,"kind":"step_started","stepId":"llm:main","attempt":1,"payload":{"name":"llm:main"},"createdAt":"..."}

event: step_completed
data: {"seq":2,"kind":"step_completed","stepId":"llm:main","attempt":1,"payload":{"output":"The three laws..."},"createdAt":"..."}

event: run_completed
data: {"seq":3,"kind":"run_completed","attempt":0,"payload":{"cost_usd":0.0004,"tokens_in":45,"tokens_out":120},"createdAt":"..."}

Dashboard view: open localhost:3001/runs to see the waterfall timeline, cost breakdown, and full event log.

What's next

Agents

System prompts, tools, visual editor, guardrails.

Connectors

Gmail, Slack, GitHub, Stripe, and 13 more.

Scheduling

Cron triggers, event triggers, email delivery.

Runtime

MicroVM isolation, durable execution, receipts.