TC Diagnose™
Know what controls are required.
Assess the workflow, suggest required controls, and promote them into Protect after human confirmation.
RUNTIME CONTROL FOR AI AGENT ACTIONS
Before an AI agent changes a system of record, TekCapitol checks the controls that matter and returns Allow, Pause, or Block.
Your existing identity, policy, approval, risk, and governance controls stay in place. Check Write™ evaluates the relevant evidence at the moment of action.
Illustrative decision. Your code still owns the write.
It returns Allow, Pause, or Block. Your code still owns the action.
npm i @tekcapitol/tc-protect-sdk
·
View on GitHub →
CRM, ERP, finance, ITSM, and databases already have identity, approval, and policy controls. An AI agent can still take a consequential action before those controls are evaluated for the write that is about to happen.
Call Check Write™ before the action. Allow, Pause, or Block. Your code still owns execution.
Call Check Write™ with the write intent. It resolves required controls, evaluates current evidence, and returns Allow, Pause, or Block.
What the agent wants to change, against which system, and with which context.
Identity, policy, approval, risk, and governance requirements for that action.
Each required control is checked against current evidence from your systems.
Allow, Pause, or Block. Your application, harness, or orchestrator proceeds or refuses.
Your application calls Check Write™ before the consequential action. Evaluation returns Allow, Pause, or Block with per-control detail. TekCapitol does not sit as a silent proxy between your systems.
Your frameworks, identity, approvals, risk, compliance, context, and systems of record stay yours.
Check Write™ resolves required controls, evaluates evidence, returns Allow, Pause, or Block, and records decision evidence (Trace).
Control resolution · evidence evaluation · decision response · Trace
Same decision pattern. Different control packs by industry.
Action: Change payment beneficiary
Identity · Delegated Authority · Transaction Policy · Risk · Approval
Approval missing → PAUSE
View industry example →Action: Update medication order
Identity · Clinical Authority · Patient Context · Medication Safety · Approval
Dosage threshold failed → BLOCK
View industry example →Action: Change production-line operating parameter
Agent Identity · Authority · Operating Context · Safety Policy · Change Approval
Operating context changed → PAUSE
View industry example →The broader TekCapitol approach for moving agent workflows into production safely. Check Write™ is the developer call inside TC Protect™.
Know what controls are required.
Assess the workflow, suggest required controls, and promote them into Protect after human confirmation.
Close the control gaps.
Generate remediation guidance and control packs.
Runtime protection for agent actions.
Check Write™ evaluates required controls and current evidence before consequential actions. Your code still owns the write.
Build and evaluate
Decide inside your trust boundary
Same decision logic. Different trust boundary.
Start with one consequential action. Call Check Write™ before the write.
Try Check Write™ → View developer docs →TekCapitol · TC360 · San Jose, CA · Runtime control for AI agent actions
Metadata-first by default. Private deployment is available when control evidence or enforcement must remain inside your environment.
Details and access models: Security & deployment.
TC Diagnose™ scores one AI agent workflow when you still need a map before wiring TC Protect™. The end state is Check Write before consequential actions. Details: Security & data handling.