Reference

Command line

The slurry CLI drives Slurry from a terminal, a CI job or a coding agent. Every command takes --json and exit codes are stable, so a check can gate a build or an agent loop.

Install

terminal

npm install -g @slurry/cli    # then: slurry help
npx @slurry/cli help          # or without installing

Node 20 or later. The package is @slurry/cli: one file, no runtime dependencies, MIT licence.

terminal

slurry login                        # approve a code in the browser
export SLURRY_KEY=slurry_acct_...   # or, in CI, a key from slurry.io/app/keys

slurry login prints a code and a link; approve it at slurry.io as a member of the organisation, and the account key is saved in an owner-only file in your user config. Add --read-only for a key that cannot change anything. A saved key is only ever sent to the API that issued it. SLURRY_KEY takes precedence, which is what CI uses. See Connecting.

Create, call, check

terminal

slurry sims create --from jira --name "Acme Jira" --wait --write-env .env
# run your system against ACME_JIRA_BASE_URL with ACME_JIRA_API_KEY, then:
slurry requests verify acme-jira --method POST --path "/rest/api/3/issue" --times 1
slurry requests validate acme-jira
slurry state verify acme-jira --collection issue --where "fields.summary=Renewal at risk" --each fields.status.name=Done --times 1

--write-env appends ACME_JIRA_BASE_URL and ACME_JIRA_API_KEY to the file instead of printing the key, so it never lands in a log or an agent transcript. New simulations accept calls from this machine only; add --allow for other addresses or --open for none. requests verify exits 1 with the near misses and the filter each one failed; requests validate exits 1 with every call the real vendor would reject.

In Git Bash on Windows, a path argument that starts with / is rewritten into a Windows path before the CLI sees it. Set MSYS_NO_PATHCONV=1; the CLI refuses such a path rather than check the wrong thing.

In CI

.github/workflows/slurry.yml

name: Slurry
on: [pull_request]
jobs:
  slurry:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      # start your system pointed at the suite's simulations here, then:
      - name: Run the Slurry suite
        env:
          SLURRY_KEY: ${{ secrets.SLURRY_ACCOUNT_KEY }}
        run: npx @slurry/cli test billing --speed 60 --junit slurry-results.xml

slurry test starts the run, waits and exits 0 only if it passed. Runs started with the CI variable set are tagged ci. If the run is still going at --timeout (default 30m) it is aborted and the simulations are put back. slurry cases pull and cases push keep a suite's cases as YAML in your repository. More in Test suites.

Coding agents

terminal

slurry agent setup          # Claude Code and Cursor: MCP server and guide
slurry agent setup --dry-run
slurry guide                # the same guide, printed

agent setup writes .mcp.json and .claude/skills/slurry/SKILL.md for Claude Code, and .cursor/mcp.json and .cursor/rules/slurry.mdc for Cursor. No key is written: the client signs in to the MCP server with OAuth. Other servers in an existing config are kept, and a file you have changed is left alone unless you pass --force.

Grade an agent run

slurry eval takes a file of up to 20 checks across simulations and exits 1 if the run does not reach the pass mark (default: every check passes). --session limits request checks to one agent run; --junit writes one test case per check.

slurry/eval.yaml

checks:
  - name: raised one issue
    type: requests
    method: POST
    path: /rest/api/3/issue
    expect: { count: 1 }
  - name: every call would pass Jira
    type: valid
  - name: the issue reached Done
    type: state
    collection: issue
    where: { fields.summary: Renewal at risk }
    each: { fields.status.name: Done }

terminal

slurry eval slurry/eval.yaml --sim acme-jira--run-42 --session run-42 --junit eval-results.xml

Commands

slurry help lists every command; slurry <command> --help shows its options and examples.

Commands	What they do
`login, logout, whoami`	Sign in through the browser, forget the key, show the plan, AI allowance and usage
`library`	Search the API library and read an integration guide
`sims`	Create, describe, wait for, pause, reset, fast-forward and delete simulations; logs, events and records; the MCP server each simulation also is
`records import`	Put your own records into a collection from JSON, JSON lines or CSV
`records defects`	Plant known data defects in a collection and save the answer key
`requests`	Find the requests your system sent, verify a call was made, validate them against the vendor spec
`state verify`	Check the records say what they should
`eval`	Grade a run: request, validity and state checks across simulations, with a pass mark
`snapshots, forks`	Snapshot, restore and diff; disposable copies for one test or agent run
`scenarios, actions`	Run scenario packs; fire one change as a named person
`rules`	Fixed responses, errors, latency, conditional matches, chaos and connection faults; export and apply them as YAML in your repository
`webhooks`	Add webhooks; find, read and replay deliveries
`keys`	Simulation keys
`suites, cases, test, runs`	Test suites as code (YAML pull and push), run them and wait, JUnit reports
`api`	Call any management API endpoint with your key
`guide, agent`	The guide for coding agents; install Slurry into Claude Code and Cursor

Output and exit codes

--json prints the API response as JSON, and errors as JSON on stderr. Error messages never carry a whole key.

Exit code	Meaning
0	OK, or the check passed
1	A check failed: verify, validate, eval, test, wait
2	Usage: an unknown command or a bad option
3	Not signed in, or the key is not allowed to do this
4	Not found
5	API or network error

SLURRY_API points the CLI at another engine (https, or http on localhost).

Next: MCP server

Install

Sign in

Create, call, check

In CI

Coding agents

Grade an agent run

Commands

Output and exit codes