Product

One environment for governed research jobs

One environment for a research job: start it, run it against approved data, review it, keep it, rank what follows. Layers below are parts of that path. They are not a menu of products.

  1. 01

    Slack or existing AI chat

    Where the job starts

    People do not migrate front ends. An analyst names a company or question in Slack or the chat client they already use.

  2. 02

    Harness

    The work loop

    Research, plan, implement, validate, checks, publish. The loop inside the environment, not a separate product brand. The loop can test the artifact and feed failures into the next run.

    Harness technical site →

  3. 03

    devBox

    The sandbox

    The job runs in a locked-down workspace with allow-listed resources only. On-prem or cloud. Native mGateway integration. Approved libraries only. Contained source control. Optional scheduling.

    devBox technical site →

  4. 04

    mGateway

    Identity and policy

    Identity on every request. Approved models and tools. mGateway is the policy layer of this environment, not a standalone offer.

    mGateway technical site →

  5. 05

    Workspace memory

    Prior work, scoped

    Artifacts land here after review. Default is that user. Share to a desk, pod, or named team is explicit. Not a firm-wide brain by accident.

  6. 06

    Queue

    What to do next

    Follow-ups and phases, ranked after work exists. The analyst picks. The system does not silently publish.

See a one-hour research run.

Book a working session. An analyst starts the job. You review the first deliverable.