kassette

inspect and fork durable agent workflow journals

TLDR

List runs in a journal storage directory

$ kassette list --storage file:[.kassette]

Show status of a run

$ kassette status [runId] --storage file:[.kassette]

Dump journal entries (optionally from an offset)

$ kassette dump [runId] --offset [N] --storage file:[.kassette]```

**Fork** a new run from a journal prefix

Use S3-compatible storage

$ kassette list --storage s3://[bucket]/[prefix]]

# SYNOPSISkassette command [options] [args]# DESCRIPTIONkassette is the CLI from @usekassette/cli for operating on journals written by the kassette durability library. Kassette records agent workflow steps in append-only JSONL journals so retries can replay completed work instead of re-running expensive LLM calls and side effects.The library embeds in your process (no separate server). The CLI helps humans and agents list, status, dump, and fork runs against filesystem or S3-backed storage. Journals are plain text suitable for jq and offline inspection.# PARAMETERSlist

List runs in the selected storage

status runId

Summarize run state (success, suspended, error, …)

dump runId [--offset N]

Print journal entries

fork srcRunId --from-offset N

Create a new run replaying a journal prefix then continuing live

--storage file:path|s3://bucket[/prefix]

Backend selection; install @usekassette/s3 for s3://

# CAVEATSThe primary product is an embeddable library; the CLI is for inspection and forking, not for running workflows by itself. s3:// requires the optional S3 package. Replay assumes workflow code still matches journal shape—use versioning and step renames carefully.# SEE ALSO[jq](/man/jq)(1)# RESOURCES

$ Source code

$ Homepage

$ Documentation