- Immediate debugging: Fetch the most recent trace of a failed or unexpected agent run with a single command.
- Bulk export for analysis: Export large numbers of traces or entire conversation threads to JSON files for offline analysis, building evaluation datasets, or regression tests.
- Terminal-based workflows: Integrate trace data into your existing tools; for example, piping output to Unix utilities like jq, or feeding traces into an AI coding assistant for automated analysis.
Installation
Setup
Set your LangSmith API key and project name:LANGSMITH_PROJECT. Replace your-project-name with the name of your LangSmith project (if it doesn’t exist, it will be created automatically on first use).
langsmith-fetch only requires the LANGSMITH_PROJECT environment variable. It automatically looks up the project UUID and saves both to ~/.langsmith-cli/config.yaml. You can also specify a project by its UUID via a CLI flag.Use with a coding agent
After you’ve installed and set uplangsmith-fetch, use your coding agent to ask questions like the following:
langsmith-fetch --help command to understand how to use the CLI and complete your request.
Find project and trace IDs
In most cases, you won’t need to find IDs manually (the CLI uses your project name and latest traces by default). However, if you want to fetch a specific item by ID, you can find them in the LangSmith UI:- Project UUID: Each project has a unique ID (UUID). You can find it in the project’s URL or by hovering over ID next to the project’s name. This UUID can be used with the
--project-uuidflag on CLI commands - Trace ID: Every trace (single execution) has an ID. In the Runs view, click on a specific run to see its Trace ID (copyable from the trace details panel). You can use
langsmith-fetch trace <trace-id>to retrieve that exact trace if you have the ID.
Usage
After installation and setup, you can use thelangsmith-fetch command to retrieve traces or threads. The general usage is:
Traces are fetched chronologically with most recent first.
Options
The commands support additional flags to filter and format the output:Output formats
The--format option controls how the fetched data is displayed:
-
pretty(default): A human-readable view with rich text formatting for easy inspection in the terminal. This format is great for quick debugging of a single trace or thread. By default:Explicitly specify the format: -
json: Well-formatted JSON output with syntax highlighting. Use this if you want to examine the raw data structure or pipe it into JSON processing tools. -
raw: Compact JSON with no extra whitespace. This is useful for piping the output to other programs (e.g., usingjqor saving directly) without extra formatting.
Fetch a trace or thread
You can fetch a single thread or trace with the ID. The command will output to the terminal by default:
--file option.
Fetch multiple
For bulk fetches of traces or threads, we recommend specifying a target directory path. Each fetched trace or thread will be saved as a separate JSON file in that folder, making it easy to browse or process them later.
traces/threads). For example, the following command will save the 10 most recent traces as JSON files in the my-traces-data directory:
--limit, the tool will output the results of the most recent, single trace to your terminal.
When sending to a directory, files will be named in the following way:
- Default: Files named by trace ID (e.g.,
3b0b15fe-1e3a-4aef-afa8-48df15879cfe.json). - Custom pattern: Use
--filename-patternwith placeholders:{trace_id}: Trace ID (default:{trace_id}.json).{index}or{idx}: Sequential number starting from 1.- Format specs supported:
{index:03d}for zero-padded numbers.
Include metadata and feedback
You can include run metadata and any feedback associated with the trace:Override the configured tracing project
To fetch traces from a different project than the one configured withLANGSMITH_PROJECT, use the --project-uuid option:
~/.langsmith-cli/config.yaml.
Export to files
You can fetch traces or full threads and export to a file:./my_threads. This is useful for exporting chat transcripts or building regression tests on multi-turn conversations. You could also use --limit with threads to fetch a specific number of recent threads, and --last-n-minutes works here as well.