Evidence-bound AI solutions for unsafe automation, missed revenue, tax overpayments, and regulated decisions.

Start with the pain—not a product name

What is costing you money, trust, or time?

Every solution below links a specific operational problem to a focused product, proof, and next step.

AI safety

Our AI could send the wrong number into a real system

Block prompt injection, altered dollars, and unverified tool arguments before payments, claims, tax, or underwriting engines execute.

Stop unsafe AI actions PrismGuard · PrismManifest · Prism-Shield
Answer quality

Our AI sounds confident even when its answer is unsupported

Catch hallucinations, digit drops, prompt attacks, and evidence gaps before release and during production.

Verify AI answers and extractions Prism-Eval · PrismShine
RAG & retrieval

Our RAG retrieves related text—not the right evidence

Control retrieval by taxonomy and intent, preserve causal context, and migrate vector databases without silently changing neighbors.

Improve RAG relevance and auditability PrismRAG · VectorPrism · PrismResonance · VectorBridge
Regulated decisions

We cannot prove why a financial or clinical decision was made

Resolve competing evidence, preserve valid and decision time, require human review, and export a decision trail that survives scrutiny.

Make regulated decisions defensible LedgerCovenant · Ledgerlock · PrismClaim · PrismClinical
Production operations

We cannot see why our agents acted—or stop them safely

Run native Python agents with a readable route ledger, replayable state, explicit gates, and an optional self-hosted operations plane.

Operate production AI with control ChorusGraph · PrismCortex · ChorusControl
Inbound growth

Website visitors leave before anyone answers or books them

Give prospects a business-specific assistant that answers questions, qualifies intent, and moves them toward bookings or sales around the clock.

Capture leads after hours Hotel · law · dental · real estate · restaurant · more
Not sure where to start?

Browse the complete solution directory

Compare every product and library by buyer problem, maturity, deployment model, evidence, and next action.

View all solutions Human-readable and machine-readable catalog
Choose how you want to engage

Build it, operate it, or deploy a finished solution

The problem map above tells you what fits. These lanes tell you how to start.

Build with open source. Install ChorusGraph, PrismGuard, PrismShine, and PrismRAG — prove glue reduction, named gates, answer verdicts, and taxonomy retrieval with free interactive demos before any sales call.

Explore product outcomes
Technical foundation

How the reliable-AI building blocks fit

For engineering teams, the open-source stack separates runtime, security, verification, retrieval, and operations so each failure boundary stays inspectable.

Agent Runtime

Native Python orchestration instead of duck-taped glue.

ChorusGraph

ChorusGraph is the Agent Runtime: own Graph engine, default-on semantic cache, Route Ledger. Not a LangGraph wrapper. Soft CTA: LEDGER. Apache-2.0 — pip install "chorusgraph==1.3.0".

Product page · Interactive demo · vs LangGraph

AI Security

Ingress firewall, money-tool authorization, runtime gateway.

PrismGuard · PrismManifest · Prism-Shield

PrismGuard decides who may speak / inject (GRADE). PrismManifest decides which dollar parameters may execute (MANIFEST). Prism-Shield enforces ACCEPT / REVIEW / REFUSE at runtime (SHIELD). Not SOC2 certified.

PrismGuard · PrismManifest · Prism-Shield

AI Verification

CI red team, then evidence-aware answer verdicts.

Prism-Eval · PrismShine

Prism-Eval fails the build on digit-drops and prompt injections before merge (EVAL). PrismShine verifies answers against a supplied preload (VERIFY); PASS means grounded in that preload, not world-true.

Prism-Eval · PrismShine · VERDICT demo

AI Retrieval

Taxonomy graphs, wavepacket memory, proven migrations.

PrismRAG · PrismResonance · VectorBridge

PrismRAG is taxonomy-controlled Graph RAG (TAXONOMY). PrismResonance is wavepacket memory for RAG. VectorBridge migrates vector databases and proves neighbour overlap. VectorPrism is a separate supporting retrieval engine, not this category.

PrismRAG · PrismResonance · VectorBridge

Operations

Optional self-hosted ops plane. Not the orchestrator.

ChorusControl

ChorusControl is Operations: Trace, Logs, Admin in your VPC. Orchestration stays on ChorusGraph. Enterprise is $1,999/month Founding (CONTROL). Open-source libraries stay $0 forever.

Product page · Pricing

View all products · Research tools · Fail-closed ops packaging · Evidence center

Production control — not another demo

Fail closed when the agent is stuck — with a ledger you can read

Building agents is crowded. Owning them in production is not: you need stop / escalate behavior, a hop ledger you can replay, and a verdict that “fluent” is not the same as grounded.

Insight ITS packages that path with open-source modules you can prove first:

  • ChorusGraph Route Ledger — persists why each hop chose (soft path LEDGER).
  • PrismManifest → Prism-Shield — Manifest seals money args; Shield enforces ACCEPT / REVIEW / REFUSE at runtime (MANIFEST · SHIELD).
  • PrismShine — evidence-aware answer verdicts on what the stack produced (VERIFY · PASS ≠ world-true).
  • ChorusControl — optional self-hosted ops plane when you need fleet Trace / Logs / Admin in your VPC (CONTROL).

This is governance for production agents — not a new agent framework and not a chatbot wrapper.

  1. Hop runs
  2. Route Ledger
  3. Shine VERIFY
  4. Continue or stop treating PASS as permission
Platform flow

How the infrastructure wires

Same products as the tree above, in production order: AI Security (Guard, Manifest) → Agent Runtime (ChorusGraph) → AI Verification (Shine) + AI Retrieval (PrismRAG). Operations (ChorusControl) is optional. Select a node for its definition — all text is in this HTML.

Not on this live hop: Prism-Eval is AI Verification in CI (digit-drops fail the build — it does not sit on the request path). Prism-Shield is AI Security at runtime (ACCEPT / REVIEW / REFUSE before engines run), beside Manifest.

PrismGuard

Self-hosted prompt-injection firewall for production LLM apps. Returns an auditable allow/block decision through named gates — not a vague score. Soft path: GRADE.

Product page · vs LLM Guard

  1. PrismGuard filters unsafe prompts with named gates.
  2. PrismManifest gates which money arguments may execute.
  3. ChorusGraph orchestrates the agent runtime and records Route Ledger hops.
  4. PrismShine verifies answer grounding (VERIFY — not world-true).
  5. PrismRAG supplies taxonomy-controlled retrieval.
  6. Prism-Eval is AI Verification in CI — not a live request hop.
  7. Prism-Shield is AI Security at runtime (ACCEPT / REVIEW / REFUSE), beside Manifest.
  8. ChorusControl optionally operates the fleet (replay-oriented ops in your VPC).

Full pack architecture & benchmarks · Prism Stack diagram

Prism Pack Family

Job-by-job pack benchmarks

Eye-to-eye FinancePackBench wins — Task + PI, prompt-injection holdout, embed tax — not a claim to “beat AWS overall.” Methods and scorecards live on the pack page.

Open Prism Pack Family Win matrix

Prism Pack Family win matrix — job-by-job benchmark summary
Prove it

Evidence center

Benchmarks, demos, packages, and deployment — disclosed, not slideware. Default view shows three high-value proof items.

Task + PI 100/100 vs LangGraph

FinancePackBench job win — eye-to-eye peers, not “beat AWS on everything.”

PI 100% vs 45% AgentCore

Prompt-injection job comparison on the published pack scorecard boundary.

Embed tax ~6× vs re-embed

PrismAPI / warm retrieve path vs naive re-embed — see pack page for methods.

Vendor-authored benchmarks. Guard → ChorusGraph → Shine, plus PrismAPI (embeds) and PrismDriver (warm retrieve). Pattern: runtime hop evidence (Route Ledger) → answer verdict (PrismShine VERIFY). Job-by-job pack wins remain on Prism Pack Family — not a claim that Shine replaces your orchestrator. See pack benchmarks · Guardrail Scorecard

Production agents fail in layers. We publish job-by-job wins on FinancePackBench — eye-to-eye peers.

Open pack benchmarks
Two lanes · Libraries free forever · One enterprise ops plane

Free open-source stack. One paid ops platform.

Prism stack libraries stay $0 forever. ChorusControl Enterprise is the self-hosted AI Operations Platform for your VPC.

Lane 1 · Open source · free forever

Prism · Chorus libraries

ChorusGraph · PrismGuard · PrismShine · PrismRAG · PrismCortex · PrismLib · …

Price
$0

No purchase required

Soft CTAs: GRADE · LEDGER · TAXONOMY · VERIFY · pip / GitHub

Explore open-source products

Libraries are free forever on GitHub and PyPI. Buy ChorusControl Enterprise ($1,999/month) only if you want the self-hosted ops dashboard in your VPC. No cold Calendly. No Packages A/B/C. ChorusControl pricing →

Start with the problem costing you the most.

Find the solution for AI risk, revenue recovery, regulated decisions, retrieval quality, production operations, or missed inbound leads—then validate it with evidence, a demo, or a focused conversation.

Frequently Asked Questions

How is Insight IT Solutions’ open-source AI infrastructure organized?

Five categories under Open-Source AI Infrastructure: Agent Runtime (ChorusGraph); AI Security (PrismGuard, PrismManifest, Prism-Shield); AI Verification (Prism-Eval, PrismShine); AI Retrieval (PrismRAG, PrismResonance, VectorBridge); Operations (ChorusControl). Supporting libraries, demos, research, and vertical website agents sit outside that tree. See the tree.

What's the difference between Vertical AI and platform libraries?

Open-Source AI Infrastructure is five categories: Agent Runtime (ChorusGraph), AI Security (PrismGuard, PrismManifest, Prism-Shield), AI Verification (Prism-Eval, PrismShine), AI Retrieval (PrismRAG, PrismResonance, VectorBridge), Operations (ChorusControl). Vertical AI (secondary) is turnkey website chat for one industry. Python Libs is a supporting hub, not a category. See outcomes.

I'm a platform engineer — where do I start?

Read the five-category tree at #outcomes. Start with free demos: ChorusGraph (Agent Runtime), PrismGuard (AI Security), PrismShine (AI Verification), PrismRAG (AI Retrieval). Then PrismManifest / Prism-Shield (Security), Prism-Eval (Verification), VectorBridge / PrismResonance (Retrieval). ChorusControl is optional Operations. Python Libs is a supporting hub, not a category. Optional free Risk Management architecture audit.

Best open-source Python agent runtimes

ChorusGraph is Insight ITS’s Apache-2.0 native Python agent runtime (pip install "chorusgraph==1.3.0") — own Graph engine, semantic cache, Route Ledger. It is a LangGraph alternative, not a wrapper. Compare: ChorusGraph vs LangGraph.

LangGraph alternatives

Use ChorusGraph when you want a native replacement runtime. Stay on LangGraph and add PrismLang as middleware — PrismLang is not a LangGraph alternative.

Prompt injection protection for production agents

PrismGuard is the self-hosted prompt-injection firewall: auditable allow/block with resolution_gate. Wire it before the LLM. Guide: prompt-injection firewall.

AI tool-call authorization

PrismManifest authorizes money tool arguments before tax, underwriting, claims, or payment engines. Runtime ACCEPT/REVIEW/REFUSE is Prism-Shield. CI red team: Prism-Eval.

AI agent Route Ledger

Route Ledger is built into ChorusGraph — it records why the agent routed, not only stdout. Demo: ChorusGraph demo.

What are Prism-Eval and Prism-Shield?

Prism-Eval is AI Verification — pre-deploy adversarial CI (digit-drops fail the build; product page · soft CTA EVAL). Prism-Shield is AI Security — runtime ACCEPT / REVIEW / REFUSE (product page · soft CTA SHIELD). Funnel: Eval → merge → Shield; PrismManifest (also AI Security) seals args underneath. PrismShine is the other Verification product (answer verdicts; PASS ≠ world-true).

What is the Python Libs hub?

Python Libs is a supporting navigation hub for six PyPI pages. It is not an infrastructure category. PrismManifest, PrismRAG, PrismResonance, and VectorBridge are first-class in the Open-Source AI Infrastructure tree (AI Security / AI Retrieval). PrismLang and PrismLib stay supporting. Eval, Shield, VectorPrism, Fabric, and Cortex have their own pages.

What is VectorPrism?

VectorPrism is a supporting intent-gated retrieval library (positional subspace multiplexing in one 1024d tensor). It is not in the AI Retrieval category — that tree lists PrismRAG, PrismResonance, and VectorBridge. It is not VectorBridge (migration). Pin: pip install "vectorprism==0.1.2". Soft CTA RECOVER. Product page.

What is PrismSQL?

PrismSQL is a hosted schema-bound NL-to-SQL compiler. Upload your contract, then POST /api/prismsql/v1/query/execute with a JWT. The model never writes SQL. The API returns sql_template + params; you run it. Companion: pip install "prismschema==0.1.0" — not a pip package itself. Product page · guide · compile vs execute.

Can I try before talking to sales?

Yes — interactive OSS demos for ChorusGraph, PrismGuard, PrismShine, PrismRAG, and Prism-Eval (no signup). For Vertical AI, request a guided live demo via contact. See Evidence center.

How much does it cost?

Prism · Chorus libraries are free forever on GitHub and PyPI. The only paid platform product is ChorusControl Enterprise ($1,999/month; soft CTA CONTROL). Request access with CONTROL; Stripe Buy when checkout is live. See ChorusControl pricing. Vertical website agents are a separate catalog (nav + shop), starting at $2,499 one-time setup plus $79–$149/month depending on industry and features. AI Agent Integration (add to an existing site) starts at $1,999 one-time + $49/month. See live prices in the product shop or ask our chat widget (answers use the same PostgreSQL catalog). All USD. Platform libraries are free forever; ChorusControl Enterprise is $1,999/month (CONTROL CTA; no fake Buy until Stripe live). Canon: foundingPlatformPricing.json. Vertical-agent one-time = deployment, KB setup, and go-live; monthly = hosting, API usage, and support. 2-year maintenance guarantee on AI Web products. Live cart prices always come from /api/products when available. Last market review: June 2025.

How much is ChorusControl?

ChorusControl Enterprise is the only paid Prism · Chorus platform product. Monthly price: $1,999/month. Soft CTA: CONTROL — request access; Stripe Buy when checkout is live. Libraries stay free forever. Price home: choruscontrol.html#pricing.

Do you serve businesses outside Orange County?

Yes — we're based in Mission Viejo, CA but deploy remotely for clients worldwide.

What makes your agents different from generic chatbots?

Industry pre-training for vertical agents, scoped knowledge bases (no off-topic hallucinations), real actions where configured, and the same Prism stack we run on this homepage — PrismLang, PrismRAG, PrismResonance, pgvector, PrismCache.

What does “fail closed” mean here?

When the agent is stuck or the answer is not grounded, you should not keep burning tools and tokens as if nothing happened. ChorusGraph’s Route Ledger makes hops inspectable; PrismShine’s VERIFY checks grounding; ChorusControl is the optional ops plane for fleet Trace/Logs. Soft paths: LEDGER · VERIFY · CONTROL. See fail-closed packaging.

Do I need ChorusGraph to use PrismShine?

No. PrismShine is standalone for verdicts. The homepage “ledger → verify” packaging is the recommended production pattern when you already run ChorusGraph (or any runtime that emits inspectable receipts).

Talk after you’ve proven the stack

Tried a demo or pip install? Tell us what you need next — custom agent/RAG/security on AWS/Azure, ChorusControl in your VPC, or a vertical website agent. Call +1 (973) 692-6919 or +1 (310) 702-7911 · email info@insightits.com · Mission Viejo, Orange County, CA · serving clients worldwide.

Schedule when you’re ready

Book a fit call after demos — not a cold calendar wall.

View all products — platform, vertical, research

Open-Source AI Infrastructure (lead)

Five categories: Agent Runtime (ChorusGraph) · AI Security (PrismGuard, PrismManifest, Prism-Shield) · AI Verification (Prism-Eval, PrismShine) · AI Retrieval (PrismRAG, PrismResonance, VectorBridge) · Operations (ChorusControl). See the tree.

Vertical AI Web Agents (secondary)

Buy a finished agent for hotel, dental, law, etc. when you want done-for-you. Priced in the product shop.

Product shop — live catalog pricing

Transparent tiers: setup + monthly from our Stripe catalog. Vertical AI agents include 2-year maintenance — a separate catalog from platform libraries ($0 forever) and ChorusControl Enterprise. Research labs under Research tools — not in this cart.

Cart bundle note — DB & payment integrations

Adding a Vertical AI Web Agent may auto-include optional Database and Payment Integration line items. Remove them in the cart if you do not need those services.

Loading products...

Contracts, secure payments & 2-year guarantee
📄

Clear & Comprehensive Contracts

Every project begins with a detailed contract that clearly outlines scope, timeline, payment terms, 2-year support, IP, and confidentiality. All contracts are reviewed and agreed upon before work begins.

🔒

Secure Payment Processing

Industry-standard secure payment methods: card processing, bank transfers, encrypted payment information, PCI DSS compliant processing, receipts, and a 30-day money-back guarantee on setup fees if not satisfied.

Your Protection Guarantee

We are a legitimate, established business with 20 years of experience in Mission Viejo, Orange County, California. All contracts are legally binding and protect both parties. Free maintenance & ticketing for 2 years from contract signing; new features beyond original scope are quoted separately.