CLI Reference

The jamjet CLI is the primary interface for managing projects, running workflows, and inspecting executions.

Installation

pip install jamjet
jamjet --version
# JamJet CLI 0.1.0

Global flags

Flag	Description
`--runtime URL`	Runtime URL (default: `http://localhost:7700`)
`--api-key KEY`	API key for hosted runtime
`--output json`	Output as JSON (for scripting)
`-v, --verbose`	Verbose output

`jamjet init`

Scaffold a new project or add JamJet to an existing one.

# New project in a new directory
jamjet init my-agent
cd my-agent

# Add to current directory (like git init)
jamjet init

Creates:

workflow.yaml — starter workflow template
jamjet.toml — project configuration
README.md — project readme

`jamjet dev`

Start the local development runtime.

jamjet dev

▶ JamJet Dev Runtime
  Port:  7700
  Mode:  local (SQLite)
  API:   http://localhost:7700

Press Ctrl+C to stop.

Flag	Description
`--port N`	Listen on port N (default: 7700)
`--db PATH`	SQLite file path (default: `.jamjet/dev.db`)
`--with-mcp-server`	Also start the MCP server
`--with-a2a-server`	Also start the A2A agent server
`--reload`	Auto-reload on workflow file changes

`jamjet validate`

Validate a workflow file without running it.

jamjet validate workflow.yaml

Valid. workflow_id=hello-agent version=0.1.0
  Nodes: 4  Edges: 3

`jamjet run`

Submit a workflow execution and wait for the result.

jamjet run workflow.yaml --input '{"query": "What is JamJet?"}'

✓ Execution started: exec_01JM4X8NKWP2
  Status: running
✓ node_completed   think   claude-haiku  512ms
✓ Execution completed.

Flag	Description
`--input JSON`	Initial state as JSON
`--input-file PATH`	Read initial state from a JSON file
`--stream`	Stream events in real time
`--wait / --no-wait`	Wait for completion (default: wait)
`--timeout N`	Timeout in seconds (default: 300)

With --stream:

jamjet run workflow.yaml --input '{"query": "Explain event sourcing"}' --stream

✓ exec_01JM5Y9... started
 → node_started    think
✓ node_completed   think   claude-haiku  489ms  (64→312 tokens)
✓ Stream complete

`jamjet inspect`

View the full state and event timeline of an execution.

jamjet inspect exec_01JM4X8NKWP2

Execution: exec_01JM4X8NKWP2
  Workflow: hello-agent v0.1.0
  Status:   completed
  Duration: 512ms

State:
  query:  "What is JamJet?"
  answer: "JamJet is an agent-native runtime..."

Events:
  09:31:00.001  execution_started
  09:31:00.012  node_started      think
  09:31:00.524  node_completed    think    claude-haiku (64 in / 312 out tokens)
  09:31:00.525  execution_completed

Token usage:
  Input:  64
  Output: 312
  Cost:   ~$0.00012

`jamjet ls`

List recent executions.

jamjet ls

ID                        Workflow         Status      Duration  Started
exec_01JM5Y9NKWP3        research-agent   completed   2.1s      2m ago
exec_01JM4X8NKWP2        hello-agent      completed   512ms     5m ago
exec_01JM3W7MKVM1        research-agent   failed      8.3s      12m ago

Flag	Description
`--workflow ID`	Filter by workflow ID
`--status STATUS`	Filter by status (running, completed, failed)
`--limit N`	Show N results (default: 20)

`jamjet resume`

Resume a waiting or failed execution.

# Resume a waiting execution (e.g., after human approval)
jamjet resume exec_01JM4X8NKWP2 --event human_approved --data '{"approved": true}'

# Retry a failed execution from the last checkpoint
jamjet resume exec_01JM3W7MKVM1 --retry

`jamjet cancel`

Cancel a running execution.

jamjet cancel exec_01JM4X8NKWP2

`jamjet tools`

Inspect and test MCP tool servers.

# List all available tools
jamjet tools list

# Call a tool directly (for testing)
jamjet tools call brave-search web_search --args '{"query": "JamJet"}'

`jamjet agents`

Inspect A2A agent cards.

# Inspect a remote agent
jamjet agents inspect https://agents.example.com/research-agent

# List locally exposed agents
jamjet agents list

`jamjet eval run`

Run an eval dataset against a workflow.

jamjet eval run dataset.jsonl \
  --workflow workflow.yaml \
  --rubric "Is the answer accurate and complete?" \
  --min-score 4 \
  --assert "len(output.answer) > 0" \
  --fail-below 0.9

Running 50 eval rows... ████████████████████ 50/50

┌─────────┬────────────┬───────┬──────────┬────────────────────┐
│ Row     │ Status     │ Score │ Latency  │ Note               │
├─────────┼────────────┼───────┼──────────┼────────────────────┤
│ row_001 │ ✓ passed   │ 4.8   │  512ms   │                    │
│ row_002 │ ✗ failed   │ 2.1   │  891ms   │ Answer too vague   │
│ ...     │ ...        │ ...   │ ...      │ ...                │
└─────────┴────────────┴───────┴──────────┴────────────────────┘

Results: 47/50 passed (94.0%) — above threshold 90.0% ✓

Flag	Description
`--workflow PATH`	Workflow file to run
`--rubric TEXT`	LLM judge rubric
`--model MODEL`	Model for LLM judge (default: claude-haiku-4-5-20251001)
`--min-score N`	Minimum score 1–5 (default: 4)
`--assert EXPR`	Python assertion (repeatable)
`--latency-ms N`	Max latency threshold
`--cost-usd N`	Max cost per row
`--concurrency N`	Parallel rows (default: 5)
`--fail-below N`	Exit code 1 if pass rate below N (default: 1.0)
`--output PATH`	Write results JSON to file

CLI Reference

On this page