Show HN: TokenSpend, the AI ROI Solution

TokenSpend, a new AI ROI solution, launched on Hacker News with a single endpoint that routes requests to Claude, GPT, Gemini, and open models while tracking usage and cost per engineer. The platform, which has already routed 1.1 million tokens across 52 models and 14 vendors, joins routed usage to Git branches, commits, and pull requests to distinguish shipped, in-flight, and unmatched spend, with sample data showing $48.2K in Claude spend of which $28.4K (59%) tied to merged pull requests.

Route every model. Read every receipt. One endpoint for Claude, GPT, Gemini and open models. Routing is free. Usage and cost stay visible by engineer. Live registry Pick the model. Keep the receipt. Claude Sonnet 5anthropic/claude-sonnet-5$2/M in1,000K context /models/anthropic/claude-sonnet-5 GPT-5.6 Lunaopenai/gpt-5.6-luna$0.2/M in1,050K context /models/openai/gpt-5.6-luna Gemini 3.5 Flashgoogle/gemini-3.5-flash$1.5/M in1,048K context /models/google/gemini-3.5-flash 52 models /models - 52 - models - 14 - vendors - 1.1M - tokens routed - — - p50 latency live registry one endpoint observed receipts 0 requests · 24h Spend review See what the spend shipped. TokenSpend joins routed usage to Git branches, commits and pull requests. Shipped, in-flight and unmatched spend stay distinct. - Aggregate by default, never an engineer leaderboard. - Confidence stays visible on every attribution. - Request history included. Zero data retention is one click away. Spend review March 2026 · Acme Engineering $48.2K Claude spend $28.4K became shipped code. $9.8K never tied to a merged PR. $28.4K 59% · Tied to a merged pull request $7.1K 15% · Open PR or inside the attribution window $9.8K 20% · No merged PR after 30 days $2.9K 6% · Other repositories or non-PR work Sample data. Quick Start produces correlation; the capture plugin upgrades selected teams to SHA-level attribution. How it works 1. Connect GitHub Securely link your repositories through the GitHub App to track what ships. 2. Track AI spend Capture Claude Code usage and cost across your fleet, automatically and zero-touch. 3. Reconcile ROI Match spend to merged pull requests and measure what your AI budget actually shipped. Enterprise features Team-level attribution Spend and shipped rate per team. Aggregate by default, never a per-engineer ranking. Automatic PR linking Each captured session resolves to the pull request it shipped in, by branch and SHA. PR 128 Checkout retry logic PR 131 Dashboard charts PR 126 Spike: vector search Security Encrypted in transit and at rest. SSO on Enterprise. SOC-2 is on the roadmap. Privacy first Token counts, timestamps, and git refs leave the machine. Your prompts, code, and transcripts never do. Audit ready Export reconciliation logs for finance and FinOps. Full visibility into how AI spend maps to shipped work. Spend over time Watch the burn, by model See daily Claude spend broken out by model version, with the percent that shipped traced on its own axis. The same chart you get on your dashboard. Spend over time Daily AI spend estimated + metered , by model version, against the spend that shipped. Scrub the plot or navigator. Click to pin a period. Arrow keys supported - Peak day - $4,673Jun 18 - Daily average - $2,028 - Shipped in range - $27,42545% Sample data.