Beta commands

platform beta is the home for commands that are live but still maturing. They work and your agent can use them; the interface may change before they settle into a stable command group.

Command	What it does
`beta jev`	Evaluate a piece of state with TypeSafe.ai’s Jev “System One” API.
`beta monid`	Use Monid, the data-provider marketplace API: discover, inspect, and run provider endpoints.

`platform beta jev`

Runs a TypeSafe.ai Jev evaluation against a piece of text or structured state. Useful for scoring output quality, routing decisions, or running judgment checks without a full model inference loop. The command authenticates through the platform proxy. No API key setup needed.

platform beta jev \
  --state "The support ticket was resolved in under 2 hours." \
  --questions '{"resolved_fast":{"type":"noul","instructions":"Was the issue resolved quickly?","criteria":"Resolved in 2 hours or less counts as fast."}}'

Arguments

Flag	Required	What it does
`--state <text>`	One of three	The state to evaluate as text.
`--state-file <path>`	One of three	Read the state from a file instead.
stdin	One of three	Pipe the state in on stdin.
`--questions <json>`	One of two	The questions object as JSON.
`--questions-file <path>`	One of two	Read the questions JSON from a file.
`--model <id>`	No	Model to evaluate with (default: `jev-latest`).
`--json`	No	Print the raw API response instead of a rendered summary.

Provide exactly one state source and exactly one questions source.

Question types

Each key in the --questions object is a label for that answer in the output. Every question has type, instructions, and criteria.

Type	What it answers	Example output
`noul`	Yes or no, with a probability	`yes (0.91)`
`choice`	Picks one from 2-255 options	`option_b (0.78 confidence)`
`score`	A score from 2-10 levels	`score 4: meets expectations (0.83 confidence)`

{
  "tone_positive": {
    "type": "noul",
    "instructions": "Is the response tone positive?",
    "criteria": "Warm, friendly, not defensive."
  },
  "quality": {
    "type": "choice",
    "instructions": "Rate the response quality.",
    "criteria": "Choose the tier that best fits.",
    "options": ["poor", "acceptable", "good", "excellent"]
  }
}

`platform beta monid`

Monid is a marketplace of data-provider endpoints the agent can call on your behalf: search the catalog, inspect an endpoint’s inputs and outputs, then run it. Calls go through the platform proxy (no API key setup), and each run bills to your workspace’s credit balance at Monid’s actual cost, with no markup. The run id, output, and cost are readable afterward.

Subcommand	What it does
`monid discover <query>`	Search the catalog for endpoints matching a query.
`monid inspect`	Read one endpoint’s schema (inputs, output shape) before spending on a run.
`monid run`	Run an endpoint. Pass inputs with `--data` (JSON body) and/or `--query-params` (JSON query values). Without `--wait` it prints the run id and returns; with `--wait` it polls until the run settles and prints the result.
`monid inspect-run`	Read a previously created run (status, output, cost). Use this to poll a `run` started without `--wait`.

platform beta monid discover "company revenue"
platform beta monid inspect --provider acme --endpoint quarterly-revenue
platform beta monid run --provider acme --endpoint quarterly-revenue \
  --data '{"ticker":"ACME"}' --wait
platform beta monid inspect-run --run <run-id>

Example

# Check whether a draft email sounds professional
echo "Hi team, attached is the Q3 report. Let me know if you have questions." | \
  platform beta jev \
    --questions-file ./checks/professional-tone.json

# Output:
# professional_tone: yes (0.94)
# concise: yes (0.88)

Get started

Agents

Talk to them

Computer

Integrations

Workspaces

`platform beta jev`

Arguments

Question types

`platform beta monid`

Example

​platform beta jev

​Arguments

​Question types

​platform beta monid

​Example

`platform beta jev`

Arguments

Question types

`platform beta monid`

Example