XORCISE

Where AI readiness is proven.

Run, evaluate and improve cyber AI agents with the XORCISE suite, coordinated in your estate by XORCISE.CONTROL. XORCISE.AI is open source at launch.

Discuss enterprise deployment Get open-source XORCISE.AI

XORCISE.CONTROL

One enterprise workspace for your agent operations.

XORCISE.CONTROL agent operations dashboard

Enterprise

Run the suite in your estate.

XORCISE.CONTROL coordinates CODE, AI, EVAL and TUNE inside your infrastructure. Connect your coding agents, CI/CD, agent runtime and MLOps, with CONNOR providing cyber intelligence.

  • Controlled access. Named accounts and scoped tokens.
  • Accountable execution. Policy checks, an audit chain and budget reserved before dispatch.
  • Measured development. Compare before and after a change, including regression or no improvement.
  • Your operating model. Run it with your team or arrange platform management with FIFTH DOMAIN.

The XORCISE apps

From agent evidence to training data.

Use the four apps independently, or bring them together through XORCISE.CONTROL. Each has a distinct job and passes evidence to the next. At launch, only XORCISE.AI is open source. CODE, EVAL, TUNE and CONTROL are commercial products; contact our team for access.

XORCISE.CODE

Commercial access

Prepare the evidence

Prepare the agent profile, context envelope, bill of materials and source alignment before a run. CODE reports evidence status; downstream evaluation makes the decision.

XORCISE.CODE: Source evidence analysis

XORCISE.AI

Open source at launch

Run the mission

Connect your agent to a contained mission and capture its actions, telemetry, artifacts and outcomes. LEAP.FROG generates the simulation; AI runs a copy from the mission store.

XORCISE.AI: Mission runs and agent traces

XORCISE.EVAL

Commercial access

Assess deployability

Assess the run against a defined operating envelope, criteria and evidence thresholds. Compare performance across runs and build an evidence-backed trust case.

XORCISE.EVAL: Evaluation scores and supporting evidence

XORCISE.TUNE

Commercial access

Prepare training data

Annotate trajectories against the mission rubric and reference solution. Export annotated JSONL or eligible SFT and DPO next-action text, with AI quality reports and provenance.

XORCISE.TUNE: T5 improvement proposal and evidence
XORCISE.AI source and docs Discuss suite access

The workflow

Connected through the evidence.

The agent profile, mission and actions stay linked as a run moves through execution, assessment and data preparation. CONTROL coordinates the work; each app retains its role.

XORCISE · the apps
The five XORCISE apps Four apps in pipeline order: XORCISE.CODE prepares the evidence boundary, XORCISE.AI runs the mission, XORCISE.EVAL assesses the run in context, XORCISE.TUNE turns it into annotated training data. Only XORCISE.AI is open source at launch. CODE, EVAL, TUNE and CONTROL are commercial products. In the Enterprise edition XORCISE.CONTROL sits between the four apps and the customer's coding agents, CI/CD, agent runtime and MLOps, and calls CONNOR for intelligence. Coding agents CI / CD Agent runtime MLOps XORCISE.CONTROL Enterprise · integrates the four apps with your infrastructure, under policy
CONNOR XORCISE.CODE the evidence boundary XORCISE.AI runs the mission XORCISE.EVAL assessment in context XORCISE.TUNE annotated data out Only XORCISE.AI is open source at launch Enterprise · the same four apps, run by XORCISE.CONTROL inside your estate
Apps Four apps: the evidence boundary, the run, the assessment, the training data. Only XORCISE.AI is open source at launch; CODE, EVAL and TUNE are commercial products.

Public XORCISE run data will be hosted on xorcise.ai.