LLM Pulse CLI

llmpulse is the native Rust command-line client for the LLM Pulse API. It queries visibility metrics, mentions, citations, sentiment, AI traffic, Search Console, recommendations, GEO Writer tasks, and webhooks from a terminal or CI job.

Install

Download the archive for your platform from GitHub Releases. Extract it, then move llmpulse or llmpulse.exe to a directory on your PATH.

To install from source with Cargo:

cargo install --git https://github.com/LLM-Pulse/llmpulse-cli --tag v2.0.0 --locked

You can also run the container locally:

docker build -t llmpulse-cli .
docker run --rm -e LLMPULSE_API_KEY llmpulse-cli ping

Each release includes archives for macOS on Intel and Apple silicon, Linux on x86-64, and Windows on x86-64. Registry and Homebrew publishing remain separate release steps.

Set up the CLI

Get an API key from the LLM Pulse dashboard, then run:

llmpulse login
llmpulse status

login verifies the key, asks you to choose a default project, and writes ~/.llmpulse/config.json. Existing configuration and profiles from the TypeScript CLI continue to work.

Common commands

# Visibility summary for the default project
llmpulse metrics summary --range 28

# Every citation in CSV format
llmpulse dimensions citations --all --format csv > citations.csv

# AI referral traffic from connected analytics
llmpulse metrics ai-traffic --range 30

# Top Google Search Console queries
llmpulse search-console queries --range 30 --per-page 50

# Launch and inspect recommendations
llmpulse recommendations launch --recommendation-type ai_visibility
llmpulse recommendations list

# Subscribe a webhook endpoint
llmpulse webhooks create \
  --event-type mention.created \
  --target-url https://example.com/webhooks/llmpulse

Command groups

Command	Purpose
status, diff, report, watch	Project summaries and terminal reports
dimensions	Projects, competitors, prompts, executions, sources, mentions, citations, locales, models, and bots
metrics	Timeseries, summary, prompt performance, share of voice, top sources, AI crawler traffic, and AI referral traffic
search-console	Search Console summary, timeseries, queries, and pages
citations	Grouped citation intelligence, citing-domain mention share, URL details, occurrences, and cached content
answers, sentiments	AI responses and sentiment records
recommendations	Recommendation runs and generation
intelligence-tasks	GEO Writer tasks
projects, prompts, competitors, collections	Project setup and resource management
annotations	Timeline annotation management
reports	Technical GEO report creation
webhooks	Subscriptions and sample payloads
export	Full JSON and CSV exports
api	Direct access to any REST path

Run llmpulse <command> --help for command-specific flags.

Output

Commands accept --format json, --format table, or --format csv. List commands also accept --page, --per-page, and --all.

llmpulse metrics summary --format json | jq '.summary'
llmpulse dimensions mentions --format table
llmpulse answers list --all --format csv > answers.csv

CSV output prefixes cells that begin with spreadsheet formula characters. This prevents exported API text from being evaluated as a formula when opened in spreadsheet software.

Configuration

Configuration lives in ~/.llmpulse/config.json. Profile files live in ~/.llmpulse/profiles/.

llmpulse config set api_key llmpulse_xxx
llmpulse config set default_project_id 123
llmpulse config use production
llmpulse config show

Resolution order is command flag, environment variable, active profile, base configuration, then the default value.

Setting	Environment variable	Default
api_key	LLMPULSE_API_KEY	None
base_url	LLMPULSE_BASE_URL	https://api.llmpulse.ai/api/v1
default_project_id	LLMPULSE_PROJECT_ID	None

The CLI accepts HTTPS API URLs. Plain HTTP is limited to localhost for local testing so an API key cannot be sent over an unencrypted remote connection.

Direct REST calls

The typed commands cover all REST operations in the OpenAPI document. The api command is useful for new parameters or automation that needs the response envelope unchanged.

llmpulse api get /metrics/summary \
  --query project_id=123 \
  --query range=28

llmpulse api post /recommendations \
  --body '{"project_id":123,"recommendation_type":"ai_visibility"}'

Paths must be relative to the configured API base URL.

Shell completions

llmpulse completions bash > ~/.local/share/bash-completion/completions/llmpulse
llmpulse completions zsh > ~/.zfunc/_llmpulse
llmpulse completions fish > ~/.config/fish/completions/llmpulse.fish

Development

cargo fmt --all -- --check
cargo clippy --all-targets --all-features -- -D warnings
cargo test
cargo build --release --locked

The API parity test reads the bundled openapi.json snapshot and fails when a REST operation is missing from the Rust manifest. In the LLM Pulse monorepo, it also checks that the snapshot matches public/openapi.json. MCP and OAuth endpoints are protocol surfaces and are intentionally excluded from CLI parity.

License

MIT. See LICENSE.