Skip to main content

From sessions to answers

Your coding agents write down everything they do. Quesma collects those session files, strips secrets on the laptop, and turns the rest into a picture your whole organization can use.

The sessions are already on the laptop

Every session is written down

Claude Code, Codex, and Cursor keep local logs of every session: prompts, tool calls, retries, and results. Quesma processes those files as they are.

Open-source collector

The collector that reads those files runs in your environment. Your engineers and security team can inspect every line.

Agents stay as they are

No proxy, no rerouting, and no change to how developers work with their agents.

Open data layer

The archive in your bucket stays yours if you stop using Quesma. The tables are Parquet and Iceberg, so you can take them with you and query them with your own engines.

Your bill shows tokens. Quesma shows the work.

Provider dashboards tell you how much your team used. They cannot tell you whether that usage fixed a bug, shipped a patch, or repeated the same failed approach for an hour.

  1. Repeated work

    The same test and context run again without a code change.

    YOUR TEAM CANStop costly retry loops.
  2. Model mismatch

    A high-cost model handles work that does not need its full capability.

    YOUR TEAM CANRoute routine work to a lower-cost model.
  3. Verified delivery

    Tested changes are separated from long sessions that ship nothing.

    YOUR TEAM CANScale the workflows that deliver.
Decisions drawn from the sessions themselves, with token totals as one input.

Your sessions stay in your cloud

Secrets, credentials, and sensitive values are removed on the developer machine, before anything is uploaded. The collector then writes directly to object storage your company controls.

YOUR DATA PLANE
  1. Agent sessionsdeveloper environment
  2. Collection + redactionbefore upload
  3. Customer storageencrypted sessions
  4. Quesma analyticsnot in the upload path
The collector writes to your storage. Quesma is not in the upload path; it decrypts a copy with your private key to build the tables. Read the architecture.

We analyze the agents. We don't score the engineers.

Quesma shows patterns across teams, repositories, and kinds of work. It does not turn those patterns into a ranking of people. Reading an individual's sessions takes permission, granted per role and never assumed.

Leadership sees the structure of the work, never a productivity score with an engineer's name attached.

  • employee AI score
  • employee token ranking
  • vibe coding compliance
  • personal slop quota

Asked before every rollout

  1. Is this a surveillance tool?

    No. The views leadership and finance get are aggregated by team, repository, and kind of work. Reading one engineer’s sessions requires an explicit, role-based grant.

  2. Does our code leave our infrastructure?

    The upload does not touch Quesma: redaction runs on the developer’s machine, and the collector writes only to object storage you own. To build the analytics, Quesma then decrypts a copy with your private key; with Enterprise BYOC nothing leaves your cloud.
    See the architecture.

  3. Do we have to change how our agents run?

    No. Developers keep their tools, settings, and models. The collector reads the local session files after the fact.

Your agents keep the logs. Start reading them.

Early access starts with teams using Claude Code, Codex, or Cursor where engineering and finance both need answers about agent spend.

Talk to the founder