← Primust · Integrations

Use the workflow you already trust.

Primust does not replace your stack. It wraps what already runs and emits the portable evidence the underlying system was never designed to produce on its own.

§ 01 — Premise

Use the workflow you already trust. Add the artifact you still need later.

Primust does not need to own your policy engine, agent framework, or enterprise platform to issue useful evidence. It wraps what already runs, records the review-relevant execution surface, and emits the same portable credential model afterward.

That makes integrations additive instead of platform-replacement projects. The catalogs below are the proof of that claim.

§ 02 — Operational model

We do not replace your stack. We sign what it does.

Integrations should be additive, predictable, and reviewable. The evidence path wraps existing systems and makes their execution portable; it does not ask a compliance team to re-platform core workflows just to satisfy a later review.

Observe first
Adapters sit beside the systems that already make decisions. They record commitments, controls, and outcomes rather than replacing the underlying engine.
Degrade, not block
The default failure mode is to mark weaker or missing evidence as a gap. Primust should not become a surprise dependency in the live transaction path.
Same artifact model
Agent frameworks, rule engines, detectors, and enterprise systems all land in the same VPEC shape: committed artifacts, controls, proof floor, surface, and gaps.
§ 02 — AI frameworks

For agents. autoinstrument().

Name
Integration path
Mode
What it does
OpenAI Agents SDK
primust
Python
Wraps Agents SDK runs, tool calls, handoffs, tracing spans.
LangGraph
primust
Python
Node-level and edge-level instrumentation. Bind graph traversal to the VPEC.
Google ADK
primust
Python
Agent Development Kit runs and tool invocations.
Microsoft Agent Framework
primust
Python
MSAF agents, skills, and orchestration.
Pydantic AI
primust
Python
Tool calls and validator decisions bound to the VPEC.
Semantic Kernel
primust
Python
SK functions, planners, and kernel invocations.
OpenTelemetry
primust
Python · JS · Go
Back OTel spans into Primust. Use for unsupported frameworks.
CrewAI
primust
Python
Crew tasks open/close per decision. Role, tool use, and handoffs recorded.
AG2 (AutoGen)
primust
Python
Multi-agent conversations instrumented. Agent-to-agent message commits.
LlamaIndex
primust
Python
Retriever, query-engine, and agent tool calls bound to the VPEC.
LiteLLM
primust
Python
Unified LLM interface. Every completion across 100+ providers recorded.
DSPy
primust
Python
DSPy modules and optimizers instrumented. Teleprompter runs captured.
Haystack
primust
Python
Pipeline components, retrievers, and generators recorded per run.
Strands
primust
Python
Strands agent tool calls and tool-use policy wrapped into the VPEC.
AWS Bedrock
primust
Python
boto3 Bedrock runtime. InvokeModel and Agent calls bound via HTTP events.
MCP
primust
Python
Model Context Protocol servers instrumented. Tool listings and calls recorded.
Custom checks
primust
Python
@record_check wraps your own governance logic, validators, or internal tools.
Custom pipeline
primust
Python
Pipeline() gives explicit run boundaries for proprietary agents and bespoke orchestration.
§ 03 — Rule engines

For decade-old policy. Wrap in place.

Name
Integration path
Mode
What it does
Open Policy Agent
primust
Rego
Decision log becomes a check execution record. Every opa eval can produce a VPEC-bound result.
Cedar
primust
Cedar
Policy evaluation results become check records with the policy pack hash.
Drools
primust
DRL · Java
Rule activations recorded. Agenda and working-memory commits attached.
IBM ODM
primust
ARL · Decision
Decision service calls wrapped. Rule flow and outcome bound to the VPEC.
AWS Cedar / AgentCore
primust
Cedar
Complementary: Cedar enforces policy, Primust proves it ran as declared.
§ 04 — Third-party detectors

For PII, prompt injection, and secrets. Same VPEC.

Drop-in subprocess or HTTP adapters. Every detector run produces a check record with the tool binary hash (local) or the recipe/model hash (cloud), so a third party can verify that the detector you say ran is the detector that actually ran.

Name
Integration path
Mode
What it does
Microsoft Presidio
check integration
local · PII/PHI
PII and PHI scanner. Runs locally — tool binary hashed, no network egress.
Pangea AI Guard
check integration
cloud API · multi
Prompt injection, PII, and malicious content. Recipes bound by hash.
Lakera Guard
check integration
cloud API · multi
Prompt injection and jailbreak detection. Request/response commits recorded.
OpenAI Moderation
check integration
cloud API · content
OpenAI's moderation endpoint. Category scores and verdicts bound to the VPEC.
ProtectAI LLM Guard
check integration
local · prompt/output
Prompt and output scanners. Per-scanner verdicts recorded as check rows.
TruffleHog
check integration
local · secrets
Secret-scanner. Detector count, verified-hit count, and rules hash captured.
HuggingFace PII
check integration
local · PII
HF transformers-based PII classifier. Entity counts and scores committed.
Semgrep
check integration
local · SAST
Static-analysis detector. Rule pack hash + per-rule verdicts recorded.
§ 05 — Financial crime & AML

3 platforms

ComplyAdvantage
Sanctions and AML screening. Every screening decision becomes a check record with the risk-bundle hash and the policy pack in effect. Python · 33 tests shipped.
NICE Actimize
Financial crime monitoring. Case disposition, SAR filing, and rule-activation logs wrapped into the VPEC. Python · 23 tests shipped.
FICO Falcon
Card fraud detection. Model scores and override decisions are bound to the committed transaction. Python · 24 tests shipped.
§ 06 — Decisioning & BRMS

4 platforms

FICO Blaze Advisor
Credit decisioning BRMS. Rule-flow execution, scoring, and adjudication outcomes bound to the VPEC — Mathematical proof floor on deterministic rules. Python + Java · 23 tests shipped.
IBM ODM
Operational Decision Manager. Decision-service calls wrapped; rule flow + outcome commit per invocation — same Mathematical ceiling. Python + Java · bundled with Credit BRMS tests.
Pega Decisioning
Customer Decision Hub. Next-best-action decisions and strategy executions recorded. Python · 27 tests shipped.
FIT Validation
Fairness, integrity, and threshold validation across decisioning runs. Python · 6 tests shipped.
§ 07 — Regulated-industry cores

3 platforms

Wolters Kluwer OneSumX
Regulatory reporting. Report lineage and calculation runs bound to the policy pack. Python · 30 tests shipped.
Guidewire
P&C insurance core. Underwriting, claims adjudication, and rating decisions captured. Python REST · 39 tests + Java in-process plugin.
Microsoft Purview
Enterprise data governance. Classification, policy, and DLP events bound to the VPEC. Python · 65 tests shipped.
§ 08 — In-process (Java / .NET)

5 platforms

Duck Creek Technologies
P&C policy admin. C# DCT Extensions — Mathematical ceiling in-process. Requires customer C# SDK + DCT Extensions framework.
Majesco CloudInsurer
L&A / P&C policy admin. C# extension framework — Mathematical proof of rating and issue logic.
Sapiens ALIS
L&A policy admin. C# extension — FINRA Reg BI / Rule 2111 annuity suitability proof without disclosing customer profile.
Sapiens DECISION
Java rules engine. Rating-factor consistency across a book — fair underwriting proof without disclosing applications.
InterSystems HealthShare
Healthcare interop. Java adapter via IRIS Java Gateway — clinical decision evidence without PHI transit.
§ 09 — Next step

Keep the systems you already run. Add the evidence layer they do not produce on their own.