@eigenpal/cli
Create, evaluate, and deploy workflows from your terminal. Agent-ready.
Install
npm i -g @eigenpal/cli
eigenpal auth login # or set EIGENPAL_API_KEY in CI
eigenpal skill install # install skill for your agent
eigenpal docs search evaluator # search release-matched docs offline
Commands
| Command | Purpose |
|---|---|
eigenpal status | Server, tenant, user, key id, workflow count. |
eigenpal init | Scaffold a workflow project. |
eigenpal auth | Profile management. |
eigenpal workflow | Workflow, evaluators, dataset, experiment. |
eigenpal run / eigenpal runs | Start runs and inspect artifacts, traces, reviews. |
eigenpal agents | Agent source, runs, datasets, triggers, and sessions. |
eigenpal skill | Install the agent skill across AI tools. |
eigenpal docs | List, search, and read bundled platform documentation. |
Use it
cd into a folder with your data, then ask your agent:
Build a workflow that extracts line items from these PDFs, judge with an LLM, seed five dataset examples, and run an experiment.
The skill teaches your agent the schemas. The platform handles versioning, evals, traces, and governance.
eigenpal init workflow extract --template pdf-extraction
eigenpal workflow validate # local validation
eigenpal run workflows.<id> --example <example> --wait # one-shot run
eigenpal workflow dataset push <id> --file ./dataset # upload dataset
eigenpal workflow experiment run <id> # batch eval
eigenpal runs list <id> --type workflow --json | jq # query as JSON
Primitives
| Primitive | Purpose |
|---|---|
| Workflow | Versioned DAG of steps. The thing that runs in prod. |
| Dataset | (input, expected_output) examples. Your ground truth. |
| Evaluator | Scorers: LLM judge, exact match, custom. |
| Experiment | Batch run of a workflow over a dataset, scored. |
Environment variables
Most users never need to set these — eigenpal auth login writes a profile to ~/.config/eigenpal/credentials.json and every command derives its config from there. Reach for env vars when you cannot run an interactive login (CI), need to switch context for one shell, or want to override a single field without editing the credentials file.
| Variable | Purpose |
|---|---|
EIGENPAL_API_KEY | Bypass the profile entirely. Set in CI to skip auth login. When set, EIGENPAL_BASE_URL (or the cloud default) is used — the active profile is not consulted, so a stale profile cannot redirect a CI run to the wrong server. |
EIGENPAL_BASE_URL | Override the server URL for one command or shell. Pairs with EIGENPAL_API_KEY to point CI at an on-prem deployment. Without EIGENPAL_API_KEY, this overrides whatever the active profile would have used. |
EIGENPAL_PROFILE | Switch the active profile for one shell without touching ~/.config/eigenpal/credentials.json. Useful for ad-hoc context switches: EIGENPAL_PROFILE=staging eigenpal status. Persistent equivalent: eigenpal auth use <name>. |
EIGENPAL_DIR | Override the workflow project directory used by init / validate. Defaults to ./eigenpal. |
Resolution precedence: command-line flags > env vars > active profile > defaults.
Support
File issues at github.com/eigenpal/cli/issues.
License
Apache-2.0. See LICENSE.