Xerg.ai

@xergHQ

Actionable economic intelligence for AI agents. ROI and attribution by team and project.

San Francisco, CA

xerg.ai

Joined March 2026

1 Following

13 Followers

Replies Reposts Media

Pinned
Xerg.ai
@xergHQ
Sep 22
Xerg is the economic control plane for AI agent fleets. One ledger, in dollars, of what your agents cost, what they wasted, and what they delivered — across every framework, department, cloud, and on-prem. Reads the logs your agents already write. Nothing to instrument.
Xerg: The Economic Control Plane for AI Agent Fleets
From xerg.ai
Xerg.ai
@xergHQ
Sep 23
Opus 5.5 costs $4/$20 per million tokens, down from $5/$25 on Opus 5. Vals AI found it also uses more tokens per task, most of all on coding. If usage goes up 25%, the savings are gone. Past that, your bill goes up. Xerg shows what each workflow costs in dollars, so you can see
Vals AI
@ValsAI
Sep 23
Replying to @ValsAI
Anthropic cut Opus 5.5’s token pricing by 20%, from $5/$25 per million input/output tokens on Opus 5 to $4/$20. However, Opus 5.5 often uses more tokens, especially on coding benchmarks (where it saw the largest gain). The discount lowers the sticker price, but it is offset by
Xerg.ai
@xergHQ
Sep 22
A tool that controls agent spend grades its own work when it reports on it. Runtime tools can stop a run. They can't prove which runs were waste. Xerg can, because it sits outside the request path and sees how every run ended.
Xerg.ai
@xergHQ
Sep 22
Three findings Xerg prices in dollars, each shipping with a fix: Retry waste → adjust the retry policy Tool loops → add a loop guard Cache thrash → correct the cache config Plus seven neutral signals that surface spend for review — never spun as waste.
Xerg.ai
@xergHQ
Sep 22
Enterprises now run dozens of agent frameworks, routers, and gateways across departments. OpenClaw. Hermes. Claude Code. OpenRouter. LiteLLM. Cloudflare AI Gateway. Xerg reads them all into one normalized ledger. Framework-neutral, outside the request path.