About Coder

Coder is the AI operating layer for enterprises. Its self-hosted, cloud-agnostic platform moves development environments and source code off developer laptops and onto centrally managed infrastructure in your cloud or on-premises. Developers and coding agents get the tools, compute, and project context they need, while platform teams manage access, enforce policies, and govern how work runs. Organizations retain their choice of models, agents, and development tools as they expand AI adoption across their developer teams. The same foundation gives builders beyond engineering a governed way to use agents to create software and automate tasks.

The Challenge: Building Visibility Into Growing AI Spend

As AI adoption accelerated across Coder, so did the need for visibility into its growing costs. Spending was spread across multiple providers and tools, making it difficult to understand total AI spend, identify cost drivers, and track changes over time.

This challenge, combined with a broader need for cloud cost visibility, led Coder to establish a dedicated FinOps function. In January 2026, Jessica Rothfuss stepped into the role of FinOps Manager, tasked with bringing greater visibility into the company's cloud infrastructure and AI token spending.

One of her first challenges was establishing a reliable baseline for what Coder was spending and where those costs were coming from.

Before Vantage, we weren't confident about what was going where or what we were even spending because it was all piecemeal across all of the different providers.

AI spend was also increasing faster than expected. As employees experimented with new models and expanded their use of AI, Coder began tracking above its original budget. Leadership needed a better way to understand those trends and plan for continued growth.

Coder's goal wasn't to restrict AI usage. Experimentation is an important part of the company's engineering culture, so its FinOps practice needed to introduce greater financial visibility without discouraging adoption.

Why Vantage

Coder moved from its initial conversations with Vantage to a signed contract in less than 60 days.

Flexibility was a major factor. Coder needed to bring together cost and usage data from several sources, some of which didn't follow standard formats. Vantage gave the team a way to work with that data without requiring Coder to standardize every source first.

For example, Coder uses an internal AI governance product whose logs weren't generated in FOCUS (FinOps Open Cost and Usage Specification) format. Jessica used Coder Agents to rewrite those logs into FOCUS, then set up GitHub Actions to send the resulting data into Vantage.

We definitely enjoyed the flexibility around it. The support from the team from our first conversation through present day has been phenomenal, and the pricing was something we were comfortable with. Our decision to move forward with Vantage was a combination of the team, the product, and the pricing.

Coder connected nearly all of its cost sources during the proof of concept, before becoming a customer. As its FinOps practice has developed, the Vantage team has continued to work closely with Jessica on new reporting, tagging, and AI cost management use cases.

We feel very supported. This is a true partnership.

Vantage gave Coder a way to centralize that AI cost data. But as a small FinOps function supporting a company moving quickly on AI, visibility alone wasn't enough. Jessica also needed an efficient way to work with the data.

That's where Vantage's agentic capabilities became a core part of Coder's FinOps workflow. Instead of relying primarily on dashboards and manually built reports, Coder uses the Vantage FinOps Agent, Vantage MCP, and Canvas to query cost data, build analyses, automate recurring reporting, and answer new questions from leadership.

Using the Vantage Agent as the Front Door to FinOps

The Vantage FinOps Agent has become Coder's primary way of working with its cost data.

Instead of starting with a dashboard or building a new report for each request, the team uses the Vantage Agent to query its cost data in Vantage directly. This is particularly useful as a small FinOps team, where requests from leadership often require pulling together data quickly.

When Coder's CEO recently asked for average Claude spend across different groups of employees, Jessica used the Vantage Agent to generate a table of users and their usage. She categorized the employees in Google Sheets and used a pivot table to produce the requested breakdown.

Normally, it's my first stop for, 'Hey, I need this information. Help me get it.'

Coder also uses the Vantage Agent to create reports, work with Virtual Tags, and determine how to structure data for other AI workflows. Jessica's manager and other frequent Vantage users at Coder have adopted the Vantage Agent as well.

Using Vantage MCP + the Vantage Agent to Automate AI Cost Reporting

As Coder's token spend increased, the team needed a more consistent way to keep leadership informed.

In monthly reviews with Coder's CEO, token spend, and the trend behind it, were often a surprise. Jessica introduced an automated AI cost newsletter every two weeks to give the CEO and his direct reports more consistent visibility into how much Coder was spending and how that spend was trending.

The newsletter includes total AI spend and token usage for the previous two weeks, spend by provider, monthly trends, model mix, top users, and updates from the FinOps team. It also helps leadership see how usage changes over time, for example, when employees begin experimenting with newly released models.

Jessica built the newsletter using the Vantage Agent, Vantage MCP, and Claude. Any MCP client can pull the same cost data; Coder used Claude to write the newsletter. She uses Vantage to organize the underlying cost and usage data, then brings that data into Claude through Vantage MCP to generate the newsletter.

The initial workflow took roughly one to two days to build. Now it only takes a few minutes to refresh. Jessica refreshes it every two weeks with the latest data and updates from the FinOps roadmap.

Every two weeks I basically say, 'Refresh this for the next two weeks. These are the roadmap changes I want you to make.'

The result is a repeatable way for Coder's FinOps team to keep leadership informed as the company's AI usage and costs continue to change.

Want to see how Jessica built it? Download the step-by-step guide to creating an AI cost newsletter with Vantage.

Using AI Cost Data for Forecasting and Budgeting

The visibility Coder has gained through its short time on Vantage is already informing its budgeting process.

With historical spend and usage trends in one place, the team can forecast AI costs under different scenarios and have more informed conversations about how much the company should expect to spend as adoption grows.

We've been able to have more budgeting conversations around, 'AI spend is a lot more than we were expecting. Are we prepared for it?'

Those conversations are beginning to extend beyond visibility and forecasting. Coder is now considering how to encourage more cost-conscious AI usage without limiting employees' ability to experiment.

One area the team is exploring is model selection. Rather than defaulting employees to the most expensive models, Coder can make a lower-cost model the default while still allowing users to choose a more expensive model when they need it.

Jessica describes the approach as "gentle governance": the security controls were in place from the start, so cost guidance can be introduced gradually without putting Coder's culture of experimentation at risk.

The team has already seen how small changes in behavior can affect spend. After an engineering manager reminded employees that they didn't need to use the most expensive model for every task, Coder saw a corresponding decline in AI spend in Vantage. Costs later increased again as new models were released, but the change gave the team an early indication of how usage guidance can affect spend.

We're trying to strike that balance of fiscal responsibility and tinkering.

Analyzing AI Economics with Canvas

Coder is also using Canvas, Vantage's AI-enabled reporting tool for custom analyses, to analyze AI usage and unit economics.

The team has built views that combine provider, model, input and output tokens, spend, and cost per million tokens. These analyses will become increasingly important as usage of Coder Agents grows.

Coder Agents usage data flows through Metronome, which Coder has connected to Vantage via the native integration. Canvas gives the team a flexible way to analyze that data, including usage by customer and model, without creating a custom report for each question.

This gives Coder a common reporting layer for both the internal AI costs it manages today and the customer usage data that will become more important as Coder Agents scales.

The Business Impact

Vantage has given Coder both the visibility to understand its growing AI spend and agentic tools to make that data easier to act on.

Leadership now has regular visibility into token spend and usage trends. Jessica can use the Vantage Agent to answer cost questions without building each analysis from scratch, use Vantage MCP to bring Vantage data into automated workflows, and use Canvas to explore AI usage and unit economics. Coder is using the resulting data to forecast AI spend and begin developing its approach to AI governance.

The company is still early in that process. Its next priority is improving allocation so it can move from understanding how much it spends to understanding which teams and projects are driving those costs.

That visibility will allow Coder to introduce more sophisticated budgeting and optimization over time while maintaining the flexibility its teams need to experiment with AI.

Looking Forward

Coder plans to continue building out its tagging and allocation strategy, starting with project- and team-level tagging. Jessica is also working with Coder's product and engineering teams to generate FOCUS-formatted logs for Coder products so customers can more easily analyze their own Coder Agents and AI governance spend.

Over time, the FinOps team expects to expand into showbacks, budgeting, and optimization as Coder's AI usage and FinOps practice mature.

For now, the focus is on establishing the visibility and processes needed to understand a rapidly changing category of spend, and using Vantage's agentic capabilities to make that work more efficient as Coder's FinOps practice scales.