Platform Governed multi-model enterprise AI

The governed, multi-model enterprise AI platform.

AgenTorQ is an enterprise AI platform, not a Salesforce-only app. AI Workmates read your data, reason over your knowledge, and take approved actions — every request routed through an intent router, grounded on live systems, run on the model you choose, and logged end to end. One engine, two editions: a standalone web workspace, or a native Salesforce managed package.

Salesforce is one supported connector — never the center. Connect it alongside every other enterprise system.

Book Demo Launch Workspace Explore Platform
Standalone or in Salesforce 600+ connectors BYOK across 14+ models RAG with citations Governed actions Audit-ready
Platform overview

One platform. Every part governed the same way.

Ten working parts share one identity, routing, policy and audit fabric. Adopt the whole platform or start with a single layer — the request flow and the governance are identical everywhere it runs. Jump to any layer below.

AI Workmates

Role-aware agents that read data, reason over knowledge, and take approved actions.

Intent Router

Grades and routes every request — data, knowledge, reasoning or action. Nothing hardcoded.

Data Bridge

Grounded reads on live records and connected apps behind SOQL and DML guards.

Knowledge Bridge

Cited RAG over folders, files and chunks — every answer traceable to a source.

ACTs · Actions

Create and update records, run Flows and send email behind approval cards.

Model Routing

Route each task across 14+ providers on your own keys, with cost control and fallback.

Connectors

600+ enterprise apps, databases, REST and GraphQL APIs, MCP servers and webhooks.

Governance

Zero-trust RBAC, PII masking, USER_MODE CRUD/FLS and a credential vault, deny-by-default.

Observability

Every prompt, model call and action on an immutable audit trail, with a live health console.

Extensibility · API

API-first: REST, webhooks, custom actions, connectors, and native Apex and LWC.
14+
AI providers, one interface
600+
Connectors across the enterprise
2
Deployment editions, one engine
100%
of prompts & actions audited
How a request flows

User → Workmate → routing → grounding → model → governed action → audit.

Every turn takes the same governed path. No request skips grounding, model selection, policy or the audit log — whether it runs in the web workspace or inside Salesforce.

1

Ask

A user asks an AI Workmate in the standalone workspace or on a Salesforce record page.

2

Route the intent

The Intent Router grades the request and routes it — data lookup, knowledge, reasoning, or a real action. Nothing hardcoded.

3

Ground on your data

The platform grounds on live records, connected apps, uploaded files and your RAG knowledge base — cited, and scoped to the user's access.

4

Select the model

Multi-model routing picks the right engine by task, cost, latency and policy — on your own keys, with automatic fallback.

5

Respond or act

It returns a cited answer, or a governed action behind an approval card — create or update records, run a Flow, send email.

6

Write to the audit log

Every prompt, model call and action is recorded to an immutable audit trail with identity, cost and outcome.

Platform architecture

One governed layer between your people and every system and model.

Users work through AI Workmates. The AgenTorQ platform routes, grounds and acts under one governance layer, then reaches your connected systems and the AI provider you choose — with Salesforce as one connected system among many.

Your data stays in your systems. Only the masked prompt reaches the model you chose, and every step is logged end to end.

Two ways to deploy · one platform

Deploy AgenTorQ anywhere. Same engine, either edition.

The AI engine, Workmates, governance, connectors and model routing are identical in both editions — the only difference is where it runs. Both are sold and supported as first-class products.

Edition · Standalone web

AgenTorQ Cloud

A standalone AI workspace that requires no CRM. Connect Salesforce, Microsoft 365, Google Workspace, Slack, GitHub, databases, REST APIs and MCP servers, and start instantly — nothing to install.

  • Universal connectivity
  • No-install start
  • Cross-functional teams
  • Centralized governance
Launch Workspace Explore Deployments
AppExchange Edition · Salesforce-native

AgenTorQ for Salesforce

A native AppExchange managed package that runs inside Salesforce — LWC, Apex, Flows, record pages and the utility bar. Admin-managed, enforced in USER_MODE, and your data stays in Salesforce.

  • Installed in your org
  • USER_MODE CRUD/FLS & sharing
  • Admin-managed access
  • Data stays in Salesforce
View AppExchange Edition Book Demo
Same AI engine · same Workmates · same governance · only the deployment differs
AI Workmates

Role-aware agents that read, reason, and act.

A Workmate is a role-aware AI agent: it reads your live data, reasons over your knowledge, and takes approved actions on your behalf. Give it a purpose, instructions and guardrails; it grounds every answer, cites its sources, and never writes without approval. Start from 218 ready templates across 16 divisions, or build your own.

  • Purpose, instructions & guardrails
  • Read → reason → act loop
  • Per-agent model & policy
  • Private per-user memory
  • Workmate teams on one task
  • 218 templates, 16 divisions
Explore Workmates by role See all Features

Sales Workmate

Briefs reps on live pipeline, flags stalled deals, drafts follow-ups and logs the next step — on approval.

Service Workmate

Answers case questions from records and knowledge with citations, then updates records on approval.

Ops Workmate

Pulls data across connectors, runs Flows and automates multi-step busywork — fully audited.

Admin Workmate

Explains schema and validation rules, drafts config and manages access with guardrails.

Developer Workmate

A developer-grade copilot for Apex, LWC, Flows and GitHub inside DevSpace Studio — propose, diff, deploy, test.

Intent Router

Every request routed on its merits, not a rulebook.

The Intent Router grades each turn and decides what it needs — a live data lookup, a knowledge answer, pure reasoning, or a real action. It re-grades on every turn, so the platform picks the right path dynamically instead of following brittle, hardcoded routing.

  • Dynamic graded routing
  • Data, knowledge, reason or act
  • Multi-step planning
  • Nothing hardcoded
Data Bridge

Grounded on live data, behind read and write guards.

The Data Bridge reads live records and connected apps so answers reflect the real state of your business — not a stale copy. In the Salesforce edition every read passes a SOQL guard and every write a DML guard, enforced in USER_MODE. Beyond CRM it reaches Gmail, Calendar, Drive and any REST or GraphQL API.

  • Live record-aware context
  • SOQL & DML guards
  • USER_MODE CRUD/FLS & sharing
  • Gmail, Calendar, Drive, REST
Knowledge Bridge · RAG

Answers you can trace to a source.

Connect folders, files and records into a retrieval layer with embeddings, chunking and citations. Workmates answer only from approved sources scoped to the user's access, and show their work — so every answer is verifiable and hallucinations have nowhere to hide. Upload PDF, Word, Excel, CSV and scanned images; OCR reads the rest.

  • Embeddings & semantic search
  • Source citations on every answer
  • Folder, file & chunk structure
  • Access-scoped retrieval
  • PDF, Word, Excel, CSV, OCR
  • Object-level include / exclude
How knowledge stays governed
ACTs · Actions

Approve & Execute: agent output becomes completed work.

ACTs turn a Workmate's proposal into a real change — create and update records, run Salesforce Flows, send email, or render an edit form. Every write is gated behind an approval card by default, destructive steps always confirm, and the result is written to the audit log. Autonomy never outruns control.

  • Create & update records
  • Run Salesforce Flows
  • Send email & edit forms
  • Approval cards, bulk approve
  • Destructive steps confirm
  • Enforced operating modes
Multi-model routing · BYOK

Route every task to the right model — on your own keys.

A stable model abstraction sits between your Workmates and the providers. Route by capability, cost, latency and policy, with automatic fallback and no vendor lock-in. Bring your own API keys and pay providers directly — AI at cost, not at a markup.

OpenAI

GPT-class models for reasoning, drafting and general enterprise workloads.

Anthropic Claude

Long-context, safety-focused models for careful reasoning and analysis.

Google Gemini

Multimodal models with strong grounding and large context windows.

xAI Grok

Fast, capable models for real-time reasoning and conversational workloads.

DeepSeek

Cost-efficient reasoning and coding models for high-volume tasks.

Mistral

Open-weight, efficient models for latency-sensitive and private deployments.

Groq

Ultra-low-latency inference for interactive, high-throughput workflows.

Self-hosted & compatible

Ollama and any OpenAI-compatible endpoint keep sensitive workloads inside your network.

14+ providers on your own keys — OpenAI, Anthropic Claude, Google Gemini, xAI Grok, DeepSeek, Mistral, Groq, SambaNova, Cerebras, Perplexity, MiniMax, AI21, Hugging Face and Ollama — with hundreds of models available via live fetch.

Bring your own keys Route by task Cost & latency control Policy & fallback Model abstraction
See all Features
Enterprise connectors

Connect anything. Salesforce is one of many.

Reach 600+ apps through the connector layer, plus REST, GraphQL, databases, MCP servers and webhooks — each with per-user OAuth and full call logging. Long-tail and enterprise systems are reachable through the connector layer, REST or MCP, not bespoke certified integrations.

CRM

Customer data and pipeline

SalesforceHubSpotDynamics 365

Productivity

Mail, calendar and chat

Microsoft 365Google WorkspaceSlackTeams

Storage

Files and documents

SharePointOneDriveGoogle DriveBox

Development

Code and delivery

GitHubGitLabJiraBitbucketAzure DevOps

Enterprise

Systems of record

ServiceNowWorkdaySAPNetSuite

Databases

SQL and warehouses

PostgresMySQLSnowflakeBigQuery

AI providers

14+ engines, your keys

OpenAIClaudeGeminiGrok

Generic & custom

Connect anything

RESTGraphQLMCP serversWebhooks
Explore Connectors
Zero-trust governance

Governed by default, so security can say yes.

Zero-trust controls wrap every request — enforced on the server, deny-by-default. Security teams get visibility and control without slowing the people doing the work.

Role-based access (RBAC)

Feature access by role and profile, deny-by-default, with per-user overrides — server-enforced.

PII masking

Named personal data is masked before any prompt leaves — only the masked prompt reaches the model.

USER_MODE CRUD/FLS

The Salesforce edition runs in USER_MODE — sharing, CRUD and field-level security always enforced.

SOQL & DML guards

Every read passes a SOQL guard and every write a DML guard before it ever touches your data.

Credential vault

Provider keys and connector credentials live in an external vault with per-user OAuth and token refresh.

Action approvals

Write actions are gated behind explicit approval cards, and destructive steps always confirm.

Model access controls

Allow or restrict specific providers and models per role, team or workflow.

Operating modes

Ask, Edit, Plan, Auto and Bypass — enforced per turn; Plan never deploys.

Data isolation

Per-user isolation keeps personal chats and tokens private; each team and tenant sees only its own data.

View Security

Compliance-ready architecture — your data stays in your systems, and only the masked prompt reaches the model you chose. Read our Privacy Policy.

Observability

Prove what the AI saw, said, and did.

Every prompt, tool call, model invocation and write action is logged with identity, cost and outcome. Turn that record into audit evidence for compliance, and into analytics for adoption, spend and model performance — with a live health console and full call logs.

  • Full prompt & action logging
  • Model, token & cost tracking
  • Per-user & per-team attribution
  • Immutable audit trail
  • Call logs & live health console
  • Exportable for compliance
Extensibility & API

API-first, and built to be extended.

Invoke Workmates and orchestration from any system over REST. Extend the platform with custom actions, REST and MCP connectors, webhooks, and native Salesforce code — on a stable, versioned model abstraction that outlives any single provider. Developers get a full copilot in DevSpace Studio for Apex, LWC, Flows and GitHub git operations.

  • REST API for agents & orchestration
  • Webhooks & event triggers
  • Custom actions & tools
  • REST & MCP connectors
  • Native Apex & LWC
  • DevSpace Studio copilot
Read the developer docs
run-workmate.js
import { agentorq } from "@agentorq/sdk";

// Run a governed AI Workmate against any connected system.
const result = await agentorq.run({
  workmate: "opportunity-risk",
  recordId: "0065g00000AbCdEfGHI",  // Salesforce, or any connector
  model: "auto",                     // multi-model routing + fallback
  governance: "enterprise"            // RBAC, PII masking, audit
});

console.log(result.summary);         // cited, grounded answer
console.log(result.actions);         // proposed, approval-gated

Put one governed AI platform over your whole enterprise.

See AgenTorQ route, ground, act and audit across Salesforce, your enterprise apps and 14+ models — in a live, tailored demo. Or open the standalone workspace and start now. Flat pricing: $20 per user / month.

Book Demo Launch Workspace View AppExchange Edition
reachout@agentorq.com