Motivation
A few weeks ago a long agent session on a premium model (Claude's Fable) quietly burned ~$100 in a couple of hours, and I only noticed the next day. That's literally what sent me looking for a tool — and how I found AgentsView. It's now my daily driver: it replaced ccusage for cost, and I lean on the cross-session search constantly to dig up old conversations (honestly better than the native one). I love it.
But the cost side is retrospective: it shows me where the money went after it's gone. The one thing that would've caught my $100 evening is a proactive nudge — and AgentsView, which already computes every number, is the natural home for it.
Proposal
An opt-in guardrail on the existing cost tracking:
- Budget thresholds per period (daily / weekly / monthly) in
config.toml. Off by default.
- Statusline state —
agentsview usage statusline gains normal → warn → critical as you approach / cross the limit (color + a small marker). No change when no budget is set.
- Optional local-first notification when a threshold is crossed — a generic webhook (and/or ntfy / a Telegram bot token). No third-party SaaS required; nothing added to the default config.
Scope
Observability-only: it reports / notifies, it never throttles or blocks an agent (AgentsView reads sessions, it doesn't control them). A natural extension of the money tracking that's already the core value. Everything stays off unless a budget is configured, and notification targets are user-supplied endpoints — consistent with AgentsView's local-first, no-telemetry stance.
Offer
Happy to implement it and open a PR if you'd accept the direction. I'd value a quick steer on two points: where you'd want the threshold config to live, and whether the notification layer belongs in scope or as a follow-up. Thanks for building this.
Motivation
A few weeks ago a long agent session on a premium model (Claude's Fable) quietly burned ~$100 in a couple of hours, and I only noticed the next day. That's literally what sent me looking for a tool — and how I found AgentsView. It's now my daily driver: it replaced ccusage for cost, and I lean on the cross-session search constantly to dig up old conversations (honestly better than the native one). I love it.
But the cost side is retrospective: it shows me where the money went after it's gone. The one thing that would've caught my $100 evening is a proactive nudge — and AgentsView, which already computes every number, is the natural home for it.
Proposal
An opt-in guardrail on the existing cost tracking:
config.toml. Off by default.agentsview usage statuslinegains normal → warn → critical as you approach / cross the limit (color + a small marker). No change when no budget is set.Scope
Observability-only: it reports / notifies, it never throttles or blocks an agent (AgentsView reads sessions, it doesn't control them). A natural extension of the money tracking that's already the core value. Everything stays off unless a budget is configured, and notification targets are user-supplied endpoints — consistent with AgentsView's local-first, no-telemetry stance.
Offer
Happy to implement it and open a PR if you'd accept the direction. I'd value a quick steer on two points: where you'd want the threshold config to live, and whether the notification layer belongs in scope or as a follow-up. Thanks for building this.