ब्लॉग पर वापस जाएँ

Phaseo Team17 नव॰ 2026पढ़ने में 3 मिनट लगेंगे

Phaseo CLI 0.4: tables, full prices and the whole management API in your terminal

Phaseo CLI 0.4 prints lists as tables, shows every model price and its conditions, adds commands for every management route, and runs the Phaseo MCP for local agents.

Phaseo CLI 0.4: tables, full prices and the whole management API in your terminal cover image

यह संपादकीय सामग्री अभी अंग्रेज़ी में उपलब्ध है। आसपास का इंटरफ़ेस अनुवादित है; अर्थ सुरक्षित रखने के लिए मूल लेख दिखाया गया है।

Phaseo CLI 0.4 is out. It makes everyday checks easier to read, shows what a model really costs, and lets you manage everything in a workspace from the terminal, not only the dashboard.

Update with:

npm install -g @phaseo/cli@latest

Lists are tables now

Keys, workspaces, models, providers, logs, activity, analytics and every management list now print as tables with headers. Long values are shortened to fit your terminal, and money uses one format everywhere, such as $1,234.50.

In terminals that support links, model IDs, providers and request IDs are clickable and open the matching page on Phaseo. phaseo logs get <id> also prints a link to the request in your dashboard.

--json output hasn't changed, so scripts that read it keep working.

Every price, not a meter count

phaseo pricing models used to say how many prices a route had. It now shows them: input, output and cache-read prices for each provider and service tier, plus the route's other charges.

$ phaseo pricing models --model openai/gpt-6-astra --provider openai
MODEL               PROVIDER  ENDPOINT       TIER               INPUT          OUTPUT    CACHE READ  OTHER
openai/gpt-6-astra  openai    text.generate  batch       from $5.00/M   from $25.00/M  from $0.50/M  Cache write text (30m) from $6.25/M, +1 more
openai/gpt-6-astra  openai    text.generate  flex        from $5.00/M   from $25.00/M  from $0.50/M  Cache write text (12h) from $12.50/M, +1 more
openai/gpt-6-astra  openai    text.generate  priority   from $20.00/M  from $100.00/M  from $2.00/M  Cache write text (30m) from $25.00/M, +1 more
openai/gpt-6-astra  openai    text.generate  standard   from $10.00/M   from $50.00/M  from $1.00/M  Cache write text (12h) from $25.00/M, +2 more
openai/gpt-6-astra  openai    text.generate  ultrafast  from $60.00/M  from $300.00/M  from $6.00/M  Cache write text (30m) from $75.00/M, +1 more

A price marked "from" changes with the size of the prompt or the time of day. phaseo pricing get <model> lists every price with the condition it applies under:

$ phaseo pricing get openai/gpt-6-astra --provider openai --tier standard
GPT-6 Astra (openai/gpt-6-astra) via openai, text.generate
METER                    PRICE  UNIT           APPLIES WHEN
Web search               $0.01  per request    always
Input text              $10.00  per 1M tokens  input tokens ≤ 272,000
Cache read text          $1.00  per 1M tokens  input tokens ≤ 272,000
Cache write text (12h)  $25.00  per 1M tokens  input tokens ≤ 272,000
Cache write text (30m)  $12.50  per 1M tokens  input tokens ≤ 272,000
Output text             $50.00  per 1M tokens  input tokens ≤ 272,000
...

Filter with --model, --provider, --endpoint, --tier and --limit. The pricing API now returns each entry's service_tier too, so the standard, flex and priority price lists for one route can be told apart.

The whole management API

The CLI now has commands for every /v1 management route, including:

  • spend-rules, budgets and department-budgets
  • byok for your own provider keys
  • rate-limits and classifiers
  • routing, observability and notifications
  • private-models, audit-events and workload-identity

Commands that delete or change something important ask you to confirm first, or need --yes in scripts. Secrets are read from stdin or the environment, never from command arguments, so they don't end up in your shell history.

If you're an owner or admin of more than one workspace, phaseo keys takes --workspace <id-or-slug>, so you can manage keys in another workspace without signing in again.

Phaseo MCP for local agents

phaseo mcp runs the Phaseo MCP server for a local coding agent, using the account you signed in with through phaseo login. There's no second sign-in.

The agent can look up models, prices, logs and usage. It's read-only unless you start it with --manage, and it can never make paid model calls. phaseo setup codex adds it to your Codex profile, and phaseo setup claude-code shows the command to add it to Claude Code.

Safer setup for coding agents

phaseo codex, phaseo claude-code, phaseo setup cline and the other coding-agent commands now ask you to approve the API key they need in your dashboard. The approval page shows exactly what will be created: the workspace, the key name, its spending limit and when it expires. If you don't approve, nothing is created.

Keys made this way expire after 90 days and have a $100 monthly spend limit. To use a key you made yourself, pass --key-stdin.

This release also adds phaseo fx for the fx coding agent, and fixes setup for Hermes and for agents whose model list came up empty.

Get started

अक्सर पूछे जाने वाले प्रश्न