Most cost tools have dashboards, budgets and alerts, and Costory has them too. What makes Costory different is how it’s built. Five ideas shape every part of the product.
1. Allocation as code
Cost only means something once it’s allocated to the way your company works: teams, products, cost centers. In Costory, that allocation is code.
- Enrichments are YAML files in git. One stable key on your resources, such as a
component label, maps to several columns: team, cost center, business capability. You generate the file from your service catalog, review it in a pull request, and publish it through the API or the Terraform provider.
- Entries are dated. When a team is renamed, last quarter’s costs keep their old owner. A reorg never rewrites the past.
- History is recomputed on every published change, so every month reads with the same rules.
- Shared costs are split by real usage. A shared database or cluster can be split each day by a usage metric, such as query time from Datadog, and totals still match the invoice.
- Virtual dimensions cover the cases where a few rules are enough: rules evaluated top to bottom, first match wins, editable in the UI or through the MCP.
- Teams reuse the same dimensions: each team has a scope that pre-filters every view to its perimeter.
It all sits on one cost model. Every bill, from your clouds, Kubernetes, data platforms, observability and LLM APIs, is normalized into the same columns and priced with the same cost metrics: Billed, Effective, List and Contracted Cost.
Read more on enrichments, dimensions, shared cost allocation, teams and Terraform.
2. Context and memory
A billing line tells you what you spent. It doesn’t tell you why, who caused it, or whether it’s normal. Billing data without context is just a number, and an answer you have to work out again every time is wasted. So Costory joins every bill with its context, and keeps what it learns.
Context: what happened around the cost.
- Telemetry and usage metrics: requests, query time, active users, tickets, from Datadog, Tsuga, Amplitude, CloudWatch, Cloud Monitoring, BigQuery, Google Sheets, S3, or your own metrics pushed over HTTP.
- Allocation from telemetry: Kubernetes, Datadog, Tsuga and dbt usage split shared costs by what each team really used.
- Events: deploys, releases, incidents, commitment purchases and provider changes, from GitHub, your CI/CD, provider feeds or the API. They sit on the cost timeline, so a change and its likely cause are on the same chart.
- Your company’s structure: owners and teams from your service catalog, through enrichments.
- Formulas: cost per active user, per ticket, per customer, so cost reads in the same units as your product.
Memory: what the system learns stays in the system.
- Definitions are shared. Dimensions, enrichments and formulas are defined once and used by every person, dashboard, alert and agent. Everyone gets the same number.
- Investigations leave a trace. When a Scout finds the cause of a change, it records it as an event on the timeline, so the next person, or the next agent, starts from it instead of starting cold.
- Corrections stick. When someone corrects a Scout in Slack (“don’t comment on staging, focus on production”), it can update its instructions, and every later run applies the correction.
- Each Scout keeps a memory of past runs, which you can open and read in the app.
This is collective knowledge. It belongs to the organization, not to one person’s chat history. Every allocation rule, annotated event and corrected instruction makes the next answer faster and more accurate. The system gets more robust as your team uses it.
Your own agent reads it too. An agent connected to the Costory MCP, in Claude, Claude Code or Cursor, reads the same events and the same Scout memory. When an engineer asks Claude why a cost moved, the answer starts from what Costory already knows.
Read more on usage metrics, events, unit economics and Scouts.
3. Every feature, at a fair price
Costory’s price follows what it costs us to run it for you, not the size of your bill.
- From 250€/month, never a percentage of your bill. Plans follow history, refresh frequency and usage. Spend bands only act as a guardrail on very large bills.
- Every plan includes every feature: allocation, dashboards, budgets, alerts, reports, unit economics, Slack, the FinOps MCP and every skill. Plans differ by history, refresh frequency, fair use, SSO and support.
- Unlimited users and data sources. No seat counting, so the whole engineering team can have access.
- Fair use, never cut off. You see your usage in the app and get a warning at 70%. Above the limit, Costory slows down, it doesn’t stop.
- Your own agent costs nothing extra. Claude, Cursor or any agent connected to the FinOps MCP doesn’t consume Ask Costory credits.
We can price it this way because every bill lands in one model on BigQuery, and processing bills costs us little.
Read more on plans and fair use.
4. Operated by humans or by agents
Costory is a product your team can run by hand, an agent can run on its own, or both. Every object is designed for both.
- The same objects everywhere. Dimensions, dashboards, alerts, reports, budgets, events and Scouts are the same objects in the web app, through the Costory MCP, through the public API, and with Terraform where a resource exists. The alert you build in the UI is the exact expression the MCP accepts and the engine evaluates.
- Agents get the context, not raw billing lines. Through the MCP, Claude, Claude Code, Cursor or VS Code query cost that’s already allocated, next to the events and usage metrics that explain it. Humans and agents read and add to the same collective knowledge.
- Skills tell agents how to do FinOps. They’re open source (Apache-2.0) and served by the MCP, so an agent runs an investigation or an allocation audit the same way every time.
- Scouts are agents that run on their own. Assign one to your reports and alerts, and it investigates what moved each time they’re delivered.
- Guardrails. Costory is read-only on your cloud. A Scout asks before it creates or changes an alert, a report or a dashboard. Recommendations go to your team or your coding agent, who decide.
Read more on the Costory MCP, skills and the API.
5. Push and pull, where engineering and product teams work
Nobody should have to remember to open a cost tool. Costory brings FinOps data to the places engineering and product teams already work, in both directions.
Pull: ask when you need it.
- In Claude, Claude Code, Cursor or VS Code, through the MCP.
- In Slack, by mentioning
@costory in any thread.
- In the web app, with Explorer and Ask Costory.
- In your own tools, with the API or an export to BigQuery. An Iceberg export (Snowflake, DuckDB and others) is in private beta.
Push: get it when it matters.
- Reports deliver a dashboard or a view to Slack, Microsoft Teams or email on a schedule, with a cost digest that ranks what moved the bill and explains it.
- Alerts notify the owner when a condition is met, including budget alerts on actual spend or the month-end forecast.
- Scouts read those reports and alerts and add the cause, the owner and the next step. If nothing is worth a ping, Slack stays quiet.
For engineering and product, not only finance. Engineers see the deploy next to the cost it caused. Product teams see cost per user or per feature, in the tools they already use.
Read more on cost reports, cost digest and alerts.
How it fits together
- One cost model brings every bill in, with the same definitions for everyone.
- Allocation as code turns that model into your company’s structure, and keeps it reviewable and stable through reorgs.
- Context explains every change, and memory keeps what was learned, so the system gets better as your team uses it.
- Every feature comes with every plan, at a price that follows usage, not your bill.
- Humans and agents work on the same objects, with the same guardrails.
- Cost goes out to the owner when it matters, and anyone can pull it from where they already work.
Last modified on November 12, 2026
