How it works
Instrument once. Understand everything.
Three steps from “where is the money going?” to “here’s the ROI report.”
Instrument in minutes
Wrap your existing OpenAI, Anthropic, or Gemini client with our zero-dependency SDK — two lines of code, no latency added. Or point your agents at our transparent proxy without touching application code at all. Import historical spend from CSV for instant baseline.
SDK Wrapper
Two lines of code. Zero latency. Works with OpenAI, Anthropic, and Gemini.
import { wrapOpenAI }
from "@cognispend/sdk";
const ai = wrapOpenAI(
new OpenAI(),
{ workflowId: "code-review" }
);Proxy Mode
Swap one URL. No code changes. Works with any language or framework.
# Before
OPENAI_BASE_URL=
https://api.openai.com/v1
# After
OPENAI_BASE_URL=
https://proxy.cognispend.com/v1Manual / CSV Import
Import historical spend or track SaaS tools like Copilot and Cursor.
date,provider,cost,team
2026-03-01,openai,4200,platform
2026-03-01,copilot,1900,frontend
2026-03-01,cursor,800,backendZero risk. All integration methods are non-blocking and never-throw. Your agents keep running even if CogniSpend is unreachable.
Attribute every dollar
Tag each AI call with a workflow, use case, team, and feature. Costs roll up across every dimension in real time. Track fixed costs — Copilot licenses, Cursor seats, self-hosted GPU — as infrastructure spend attributable to the same use cases.
Source
Agent Runs
customer-support
code-review
sales-prospecting
Every dollar gets a home. No more mysterious line items. Unattributed spend triggers an automatic insight alert.
Measure ROI, not just cost
Define what outcomes each AI workflow produces — PRs reviewed, tickets resolved, revenue attributed. CogniSpend calculates cost-per-outcome and ROI ratio per use case, updated live as events flow in. Finally, a number you can put in front of your CFO.
Customer Support Triage
Sales Prospecting Pipeline
Stay ahead of problems. Set budget alerts, get Slack notifications on cost spikes, and catch unattributed spend automatically.
Architecture
How the data flows
Non-blocking instrumentation → real-time processing → instant visibility.
Your Agents
OpenAI, Anthropic, Gemini
SDK / Proxy
Non-blocking capture
CogniSpend
Attribution & analytics
Dashboard
Cost, ROI, alerts
Know what your AI is costing. Know if it's worth it.
Set up in under ten minutes. No infrastructure changes required. Start with your existing OpenAI or Anthropic client.
Free plan available. No credit card required to start.