Give the agent a key that stops
Agent runs are temporary. Provider keys are not. Before a research sprint, a loop, or an experiment, issue a Till pass with a required request ceiling so the job can finish and then cannot keep going.
Bound the job, not the provider account.
-
01
One agent, one operating envelope
Every Till key requires a request ceiling. Add optional token, estimated spend, expiry, and source IP controls when the run needs a tighter bound.
-
02
Loops and retries are still requests
A ceiling is the operation limit for that run. It is not a rate limit and not a substitute for provider billing alerts.
-
03
Revoke one agent
Pull that run's access without rotating every provider connection. The Till key is still a bearer secret; the difference is that it can end.
-
04
Pro is the plan for a team that hands out agent keys
$59/month, 25 scoped keys, 25,000 monthly activations (plan limits). Provider inference is billed by the provider. Till charges a platform subscription.
Till is a hosted control layer for AI provider credentials. Connect provider accounts once. Each workload gets a disposable Till key with a required request ceiling and optional token, estimated-spend, expiry, and IP controls across 12 providers. Till is not a zero-knowledge vault, prompt firewall, secrets-manager replacement, compliance certification, or a replacement for provider billing or IAM.
These are plan limits, not usage claims.
Pro
Scale
Public anonymous signup is closed. New accounts are approved and provisioned manually. Existing customers can open the dashboard. Compare the full ledger on pricing.
Give it enough. Not everything.
Request Pro beta access. Till is a controlled beta; accounts are onboarded manually. Do not start from self-serve checkout.