Your terminal.
A little more capable.
One account. Your keys. Clear limits.
Start with your Google account.
Sign in, request access, and connect your terminal.
New
accounts need approval before using models.
Almost connected.
This gateway is installed, but Google sign-in has not been configured. The owner needs to finish backend setup before accounts can be created.
01 / Your access
Your account is awaiting approval or has been disabled. You can prepare a key now; model requests stay blocked until access is enabled.
Ends all your evlapi browser sessions, including this one. API keys and running requests are unchanged; revoke keys separately if needed.
02 / API keys
select a key to manage its accessCreating or revealing a key requires a Google sign-in within the last 5 minutes. Continue with Google. Google may reuse your existing Google session; this is not a password or MFA challenge.
03 / Call the API
Codex runs on the serverCreate a session, then send only each new message to its turns endpoint. Use the same API key throughout. Choose session, stateless, or both for each key above.
Read the reply from output_text. Sessions expire after
10 idle minutes, one hour total, or a server restart. A lost session
returns an error; it never silently forgets your conversation.
Keep keys on your server, out of browser code and source control.
This first adapter accepts text only (64 KiB per message). Images, files, client tools, streaming, and the OpenAI Responses protocol are not supported yet. Existing Roblox clients have not switched.
Manage API key
Creating or revealing a key requires a Google sign-in within the last 5 minutes. Continue with Google. Google may reuse your existing Google session; this is not a password or MFA challenge.
Keep keys on your server, out of browser code and source control.
Create or select a key to see its details.
Account & global controls
administrator onlyReference-dollar and token allowances are shared across each person’s keys. These are access limits, not subscription charges. Changes apply to new requests; requests already running can finish.
Saving administrator policies requires a Google sign-in within the last 5 minutes. Continue with Google. Google may reuse its existing session; this is not a fresh password or MFA challenge.
Global model catalog & pricing — all accounts and keys
No models are enabled by default. Verify the exact API model ID, maximum context, and current OpenAI prices before enabling one. Use the highest applicable input/output rate for long contexts.
People & permissions — account allowances shared across keys
Showing the newest 200 accounts.
Pending reservations
Active requests appear here too. If a stream disconnects or billing is uncertain, its reservation stays held. Reconcile with OpenAI usage before adjusting it in the database; there is no automatic refund.
Recent policy changes
5-hour and 7-day budgets roll continuously. Each request temporarily reserves a configured token allowance, then settles from CLI-reported usage valued at reference API prices. The CLI does not enforce a hard output-token cap; in-flight usage can exceed a reservation. This is not a payment or subscription charge.