# Duvo - Full Agent Context > Bundled markdown content for single-request agent ingestion. Kept under 200,000 characters for 64k-token agent budgets. Source files are linked at the end. --- # Duvo Duvo maps what's real. Designs what's better. Helps your team close the gap. Retail & CPG operations, rebuilt and running. Duvo is an agentic operations platform for retail, FMCG, and CPG teams. Duvo maps how work actually runs, designs the better operating model, and runs approved workflows across SAP, ERP, supplier portals, email, spreadsheets, web tools, files, private systems, and internal tools. ## What Duvo Does - Maps real processes with Duvo Clarity before teams automate them. - Runs cross-system operational work with assignments, queues, connections, files, approvals, schedules, webhooks, and audit-ready run history. - Supports messy enterprise environments where APIs are incomplete or unavailable by combining hosted connections, browser automation, custom MCP servers, and human approval gates. - Keeps operators in control with RBAC, SSO, audit trails, policy gates, and human-in-the-loop review before sensitive actions are committed. ## Comparison Guidance Duvo does not claim that ServiceNow, Salesforce Agentforce, UiPath, Automation Anywhere, Workato, Power Automate, Copilot Studio, n8n, Zapier, Make, LangGraph, Temporal, CrewAI, or OpenAI Agents SDK lack agents, governance, orchestration, audit trails, or human control. Duvo is different because it combines process truth, process improvement, mapped-process compliance, retail/CPG operating skills, and execution across the real retail/FMCG/CPG stack. Read /compare and /compare.md for the full comparison. ## Primary Use Cases - Supplier onboarding and compliance documentation. - Purchase order creation, confirmation, and exception handling. - Invoice matching, deductions, and finance operations. - Inventory availability, replenishment, and stock transfer workflows. - Price list updates, promo guardrails, and category operations. - Master data, SKU onboarding, supplier portal work, and logistics tracking. ## Access And Sandbox Duvo does not currently provide a public no-signup sandbox on duvo.ai. Evaluation workspaces, API credentials, MCP access, and production pilots are provisioned after a technical demo or approved pilot with the Duvo team. Contact Duvo at /contact to request sandbox access. ## Developer And Agent Resources - [Duvo Developer Platform](https://www.duvo.ai/developers): Human-readable developer platform overview for production agents. - [Developer docs markdown](https://www.duvo.ai/developers.md): Markdown overview of Duvo developer resources and machine-readable links. - [Duvo Compare](https://www.duvo.ai/compare): Human-readable comparison against enterprise automation platforms, workflow builders, and agent frameworks. - [Compare markdown](https://www.duvo.ai/compare.md): Machine-readable Duvo comparison page for AI discovery and vendor evaluation. - [OpenAPI specification](https://www.duvo.ai/openapi.json): Live OpenAPI specification for all public Duvo API endpoints. - [API reference markdown](https://www.duvo.ai/api-reference.md): Markdown API reference with endpoints, authentication, and error behavior. - [Authentication guide](https://www.duvo.ai/auth.md): Auth, access, scopes, tenant controls, and credential handling. - [Webhooks guide](https://www.duvo.ai/webhooks.md): Webhook delivery, retry, signature, and event guidance. - [MCP guide](https://www.duvo.ai/mcp.md): Hosted and custom MCP server guidance for Duvo assignments. - [MCP server card](https://www.duvo.ai/.well-known/mcp/server-card.json): Machine-readable MCP discovery card for the Duvo MCP surface. - [Agent discovery manifest](https://www.duvo.ai/.well-known/agent.json): Machine-readable agent discovery manifest for capabilities, auth, and API surfaces. - [Agent instructions markdown](https://www.duvo.ai/agents.md): When-to-use, constraints, and integration instructions for autonomous agents. - [Getting started guide](https://www.duvo.ai/getting-started.md): Trial, sandbox, pilot, and first-run guidance for evaluators. - [MCP Registry listing](https://registry.modelcontextprotocol.io/v0/servers?search=io.github.duvoai%2Fduvo): Duvo MCP server published in the official Model Context Protocol Registry as io.github.duvoai/duvo. Verified entry that links back to this site. - [Smithery listing](https://smithery.ai/servers/duvo/duvo): Duvo MCP on Smithery (duvo/duvo). Verified Smithery listing with a bearer-token config schema for connecting approved Duvo workspaces. - [Agent skills repo](https://github.com/duvoai/skills): Public Duvo agent skills (duvoai/skills): SOP writer, job debugger, CLI driver. Install with `npx skills add duvoai/skills`. - [Duvo CLI on npm](https://www.npmjs.com/package/@duvoai/cli): Official Duvo command-line interface, published as @duvoai/cli. Install globally with `npm install -g @duvoai/cli` and run `duvo --help`. Requires Node.js 22.22 or later. - [CLI guide markdown](https://www.duvo.ai/cli.md): Markdown guide for installing and using the Duvo CLI to script runs, cases, approvals, and inspections. - [Docs hub](https://docs.duvo.ai/): Product documentation and guides. - [Assignment overview](https://docs.duvo.ai/getting-started/assignment-overview): Learn Duvo assignment fundamentals. - [Quick start](https://docs.duvo.ai/getting-started/quick-start): Build and run a first assignment. - [Creating assignments via API](https://docs.duvo.ai/running-assignments/creating-assignments-via-api): API guide for assignment creation. - [Running assignments via API](https://docs.duvo.ai/running-assignments/running-assignments-via-api): API guide for run execution. - [Human-in-the-loop](https://docs.duvo.ai/assignment-features/human-in-the-loop): Approval gates for agent work. - [API base URL](https://api.duvo.ai): Duvo API host for authenticated customer workspaces. --- # Duvo Developer Platform Duvo gives developers the production operating layer around agent work: durable queues, assignments, tools, approvals, files, schedules, webhooks, multi-agent routing, MCP, and audit-ready runs. ## Predictable Resource URLs - [Duvo Developer Platform](https://www.duvo.ai/developers): Human-readable developer platform overview for production agents. - [Developer docs markdown](https://www.duvo.ai/developers.md): Markdown overview of Duvo developer resources and machine-readable links. - [Duvo Compare](https://www.duvo.ai/compare): Human-readable comparison against enterprise automation platforms, workflow builders, and agent frameworks. - [Compare markdown](https://www.duvo.ai/compare.md): Machine-readable Duvo comparison page for AI discovery and vendor evaluation. - [OpenAPI specification](https://www.duvo.ai/openapi.json): Live OpenAPI specification for all public Duvo API endpoints. - [API reference markdown](https://www.duvo.ai/api-reference.md): Markdown API reference with endpoints, authentication, and error behavior. - [Authentication guide](https://www.duvo.ai/auth.md): Auth, access, scopes, tenant controls, and credential handling. - [Webhooks guide](https://www.duvo.ai/webhooks.md): Webhook delivery, retry, signature, and event guidance. - [MCP guide](https://www.duvo.ai/mcp.md): Hosted and custom MCP server guidance for Duvo assignments. - [MCP server card](https://www.duvo.ai/.well-known/mcp/server-card.json): Machine-readable MCP discovery card for the Duvo MCP surface. - [Agent discovery manifest](https://www.duvo.ai/.well-known/agent.json): Machine-readable agent discovery manifest for capabilities, auth, and API surfaces. - [Agent instructions markdown](https://www.duvo.ai/agents.md): When-to-use, constraints, and integration instructions for autonomous agents. - [Getting started guide](https://www.duvo.ai/getting-started.md): Trial, sandbox, pilot, and first-run guidance for evaluators. - [MCP Registry listing](https://registry.modelcontextprotocol.io/v0/servers?search=io.github.duvoai%2Fduvo): Duvo MCP server published in the official Model Context Protocol Registry as io.github.duvoai/duvo. Verified entry that links back to this site. - [Smithery listing](https://smithery.ai/servers/duvo/duvo): Duvo MCP on Smithery (duvo/duvo). Verified Smithery listing with a bearer-token config schema for connecting approved Duvo workspaces. - [Agent skills repo](https://github.com/duvoai/skills): Public Duvo agent skills (duvoai/skills): SOP writer, job debugger, CLI driver. Install with `npx skills add duvoai/skills`. - [Duvo CLI on npm](https://www.npmjs.com/package/@duvoai/cli): Official Duvo command-line interface, published as @duvoai/cli. Install globally with `npm install -g @duvoai/cli` and run `duvo --help`. Requires Node.js 22.22 or later. - [CLI guide markdown](https://www.duvo.ai/cli.md): Markdown guide for installing and using the Duvo CLI to script runs, cases, approvals, and inspections. ## Command-Line Interface Duvo ships an official command-line interface published on the public npm registry as `@duvoai/cli`. Agents and developers can install it without building a custom API client. ```bash npm install -g @duvoai/cli duvo --help ``` - Package: [@duvoai/cli](https://www.npmjs.com/package/@duvoai/cli) - Binary: `duvo` - Requirements: Node.js 22.22 or later - CLI documentation: https://docs.duvo.ai/cli - Authentication: stored API key, configured via `duvo login` or the `DUVO_API_KEY` environment variable (with optional `DUVO_PROFILE` and `DUVO_API_BASE_URL`). Common commands: ```bash duvo runs start --agent --json duvo runs respond --approve duvo cases create --queue --title "case title" ``` See [/cli.md](/cli.md) for the full agent-readable CLI guide. ## Agent Integration Surfaces - OpenAPI: /openapi.json and /.well-known/openapi.json. - Markdown API reference: /api-reference.md. - Official CLI: `npm install -g @duvoai/cli` (binary: `duvo`, npm: https://www.npmjs.com/package/@duvoai/cli, guide: /cli.md). - Authentication guide: /auth.md. - Hosted and custom MCP guide: /mcp.md. - Webhook guide: /webhooks.md. - Pricing and sandbox access: /pricing.md. - AI crawler file: /llms.txt and /.well-known/llms.txt. - Markdown homepage fallback: /index.md. - Competitive comparison: /compare and /compare.md. - Agent discovery file: /.well-known/agent.json. - A2A agent card: /.well-known/agent-card.json. - Agent manifest aliases: /.well-known/agents.json and /.well-known/ai-agent.json. - OpenAI plugin-style manifest: /.well-known/ai-plugin.json. - MCP server card: /.well-known/mcp/server-card.json. - MCP manifest alias: /.well-known/mcp.json. - Agent instructions: /agents.md and /agent.md. - Live docs hub: https://docs.duvo.ai/. - Live API docs: https://docs.duvo.ai/running-assignments/creating-assignments-via-api and https://docs.duvo.ai/running-assignments/running-assignments-via-api. ## Constraints - API, MCP, webhook, and sandbox access require a provisioned Duvo workspace. - Agents cannot bypass workspace RBAC, SSO, customer policy gates, or human approvals. - Duvo is not a replacement for an ERP, MDM, WMS, TMS, or planning system; it operates across those systems and writes back governed outcomes. - Production use cases should define approval thresholds, audit requirements, data residency needs, and allowed tools before launch. --- # Duvo API Reference Base URL: https://api.duvo.ai The Duvo API lets approved customer workspaces start and inspect agent runs, create queue cases, respond to human approval gates, manage sandboxed execution, and connect external systems. API access requires a provisioned workspace. ## Machine-Readable Specification - OpenAPI: https://www.duvo.ai/openapi.json - Well-known OpenAPI: https://www.duvo.ai/.well-known/openapi.json - These routes proxy the live Duvo Public API specification and are the source of truth for the full endpoint list, schemas, operation IDs, security schemes, and response models. ## Authentication Use workspace-issued bearer credentials. Enterprise workspaces can use scoped OAuth client credentials where enabled. See /auth.md for scopes, token handling, and access constraints. ## Selected Endpoint Families This markdown file is a quick orientation, not the exhaustive API reference. Use the OpenAPI specification above before generating clients or calling endpoints. - POST /v1/agents: Create or register an assignment-backed agent worker. - POST /v1/runs: Start an agent run from an assignment, queue case, file, webhook, schedule, or external application. - GET /v1/runs/{run_id}: Inspect run status, owner, timestamps, and outcome. - GET /v1/runs/{run_id}/messages: Read run messages, tool calls, files, approvals, and audit events. - POST /v1/runs/{run_id}/human-requests/{request_id}/respond: Approve, reject, or answer a human gate. - POST /v1/queues/{queue_id}/cases: Create a durable case for an assignment or queue consumer. - POST /v1/sandboxes: Request a governed sandbox for browser, file, or desktop execution. - POST /v1/mcp: Connect to the hosted MCP endpoint for approved workspaces. ## Errors API errors should return JSON with code, message, request_id, and optional details. Clients should treat 401 as missing/expired credentials, 403 as insufficient workspace scope or policy, 404 as missing workspace-owned resource, 409 as state conflict, and 429 as a retryable rate limit response when Retry-After is present. ## Agent Retry Guidance - Mutating requests should include an Idempotency-Key header when retry safety matters. - 202 Accepted responses represent asynchronous work. Poll GET /v1/runs/{run_id} until the run reaches completed, failed, or canceled. - Clients should honor RateLimit-Limit, RateLimit-Remaining, RateLimit-Reset, Retry-After, Deprecation, and Sunset headers when present. --- # Duvo Auth And Access Duvo access is workspace-scoped. Agents, API clients, MCP servers, and webhooks inherit the policies, approvals, and permissions configured for the customer workspace. ## Discover - Protected resource metadata: https://api.duvo.ai/.well-known/oauth-protected-resource. - OAuth authorization server metadata: https://login.duvo.ai/.well-known/oauth-authorization-server. - Agent-readable auth guide: https://www.duvo.ai/auth.md. - OpenAPI security schemes: https://www.duvo.ai/openapi.json under components.securitySchemes. ## Pick A Method - Use workspace-issued bearer API tokens for approved workspace API access. - Use scoped OAuth client credentials for enterprise workspaces where Duvo has enabled OAuth for the customer tenant. - Use SSO and RBAC for human users in the Duvo application. - Use managed OAuth/API connections for third-party business systems such as ERP, CRM, email, spreadsheets, supplier portals, and collaboration tools. - Do not attempt public dynamic agent registration unless Duvo has explicitly provisioned that capability for the workspace. ## Register Public self-serve dynamic agent registration is not currently available on duvo.ai. Agent clients, API clients, OAuth client credentials, MCP access, and sandbox workspaces are registered by Duvo during an approved customer or pilot onboarding process. To request registration, contact Duvo at /contact with the workspace name, workflow category, redirect URI if applicable, requested scopes, expected systems, data residency needs, and the agent operator or vendor identity. ## Claim After Duvo provisions access, the workspace admin or agent operator receives the allowed auth method, client identifier or token handling instructions, allowed scopes, and the API/MCP surfaces approved for that workspace. Credentials should be stored in a secret manager and never embedded in prompts, logs, client-side code, or public repositories. ## Use Credential - Send workspace bearer credentials as Authorization: Bearer . - For enabled OAuth client-credential tenants, request a token from the configured authorization server and use the issued access token as Authorization: Bearer . - Include Idempotency-Key on mutating API requests when retry safety matters. - Treat 401 as missing, expired, malformed, or revoked credentials. - Treat 403 as insufficient scope, workspace policy denial, RBAC denial, or human approval requirement. ## Errors - invalid_token: refresh, rotate, or request a newly issued credential. - insufficient_scope: request the narrow missing scope through a workspace admin or Duvo contact. - workspace_policy_denied: the workspace policy or approval gate blocks the action. - approval_required: pause the agent run and wait for an authorized human response. - rate_limited: honor Retry-After and retry with the same Idempotency-Key when the request is safe to retry. ## Revocation Workspace admins can revoke or rotate issued API tokens, OAuth clients, managed connections, webhook secrets, and MCP access through the approved Duvo workspace process. Agents must stop using revoked credentials immediately, clear cached tokens, and require a newly provisioned credential before resuming privileged work. ## Supported Auth Models - Bearer API tokens for approved workspace API access. - Scoped OAuth client credentials for enterprise workspaces where enabled. - SSO and RBAC for human users in the Duvo application. - Managed OAuth/API connections for third-party business systems such as ERP, CRM, email, spreadsheets, supplier portals, and collaboration tools. ## Agent Constraints - Agents cannot see raw third-party passwords when credentials are stored in managed connections. - Agents cannot bypass human approval gates, workspace RBAC, SSO, audit logging, or customer policy. - Sensitive write actions should be scoped and routed through human-in-the-loop approval where required. - API and MCP access should use the least scopes needed for the workflow. ## Common Scopes - runs:read - inspect run status, messages, tool calls, and audit events. - runs:write - start, pause, resume, or replay runs. - queues:write - create and update queue cases. - approvals:write - respond to human approval requests. - files:read - read files attached to authorized runs or cases. - sandboxes:write - request governed browser, file, or desktop execution. - webhooks:manage - manage webhook subscriptions for a workspace. --- # Duvo Webhooks Duvo webhooks let approved workspaces subscribe to run, case, assignment, approval, and integration events. Webhook subscriptions require workspace admin access. ## Event Categories - run.started, run.paused, run.completed, run.failed - run.message.created, run.tool_call.completed, run.file.created - human_request.created, human_request.responded - queue.case.created, queue.case.claimed, queue.case.completed - assignment.build.published - connection.updated, connection.action_required ## Delivery Expectations - Payloads are JSON. - Delivery should be retried for transient 5xx responses. - Consumers should use idempotency keys or event IDs because retries can duplicate events. - Production endpoints should verify HMAC signatures and reject stale timestamps. - Webhook secrets should be rotated when ownership or integration scope changes. ## Signature Verification Webhook consumers should verify the Duvo signature header against the raw request body with the workspace webhook secret before acting on an event. Reject payloads with missing signatures, stale timestamps, or mismatched HMAC values. --- # Duvo MCP Duvo supports MCP as an agent integration surface for approved workspaces. Teams can use hosted Duvo MCP tools or connect custom MCP servers that expose proprietary systems, data sources, and workflows to assignments. ## Hosted MCP - Endpoint: https://api.duvo.ai/v1/mcp - Access: workspace-issued credentials and scopes. - Use cases: start runs, create cases, inspect approvals, retrieve run messages, and expose Duvo tools to compatible agents. ## Custom MCP Servers - Use custom MCP servers to expose internal systems, domain-specific tools, and proprietary workflows. - Keep tool descriptions specific, scoped, and auditable. - Route sensitive actions through human approvals or policy gates when needed. - Treat MCP tools as governed production capabilities, not arbitrary shell access. ## Discovery - Server card: https://www.duvo.ai/.well-known/mcp/server-card.json - Well-known MCP descriptor: https://www.duvo.ai/.well-known/mcp - MCP Registry listing: https://registry.modelcontextprotocol.io/v0/servers?search=io.github.duvoai/duvo - Registry server name: io.github.duvoai/duvo (GitHub-verified). A domain-verified ai.duvo/duvo entry is staged via /.well-known/mcp-registry-auth. - Smithery listing: https://smithery.ai/servers/duvo/duvo (duvo/duvo, verified, bearer-token config schema). - Agent skills repo: https://github.com/duvoai/skills (Duvo SOP writer, job debugger, CLI driver). --- # Duvo CLI Duvo publishes an official command-line interface on the public npm registry as `@duvoai/cli`. The CLI lets agents, evaluators, and developers script Duvo runs, cases, approvals, files, and inspections without building a custom API client from scratch. ## Install ```bash npm install -g @duvoai/cli duvo --version duvo --help ``` - Package: [@duvoai/cli](https://www.npmjs.com/package/@duvoai/cli) - Binary: `duvo` - Runtime: Node.js 22.22 or later - Distribution: npm (`npm`, `pnpm`, `yarn`, `bun`, `npx` are all supported). Alternate install commands: ```bash pnpm add -g @duvoai/cli yarn global add @duvoai/cli npx @duvoai/cli --help ``` ## Authentication The CLI stores a workspace API key in a named profile. Configure once with `duvo login`, or set the `DUVO_API_KEY` environment variable for non-interactive agent use. Use `DUVO_PROFILE` to select a named profile and `DUVO_API_BASE_URL` to override the API host. See [/auth.md](/auth.md) for the full authentication contract. ```bash duvo login export DUVO_API_KEY="" export DUVO_PROFILE="" duvo whoami ``` ## Common Commands ```bash duvo runs start --agent --json duvo runs get duvo runs messages duvo runs respond --approve duvo cases list --queue duvo cases create --queue --title "case title" duvo cases create --queue --from-file cases.json duvo sandboxes create duvo api GET /v1/runs/ ``` ## When To Use The CLI - Quick evaluation of a Duvo workspace without writing API client code. - Shell scripts, cron jobs, and CI/CD pipelines that need to start, inspect, or respond to runs. - Local debugging of assignments, queue cases, and approvals. - Agent and operator workflows that prefer a stable command surface over raw HTTP. ## Constraints - The CLI calls the same workspace-scoped API documented in [/api-reference.md](/api-reference.md) and [/openapi.json](/openapi.json); it cannot bypass workspace RBAC, SSO, policy gates, or human approvals. - API access requires a provisioned Duvo workspace and a workspace-issued bearer token. See [/auth.md](/auth.md). - The CLI is published as an ES module and requires a recent Node.js LTS release. ## Related Resources - CLI documentation hub: https://docs.duvo.ai/cli - Developer platform overview: [/developers.md](/developers.md) - API reference: [/api-reference.md](/api-reference.md) - OpenAPI specification: [/openapi.json](/openapi.json) - Authentication guide: [/auth.md](/auth.md) - MCP guide: [/mcp.md](/mcp.md) - Webhooks guide: [/webhooks.md](/webhooks.md) - Getting started: [/getting-started.md](/getting-started.md) --- # Duvo Getting Started This guide answers the first questions an agent or evaluator needs before using Duvo. ## Can I Try Duvo For Free Without Signup? No. Duvo does not currently offer a public no-signup sandbox on duvo.ai. Duvo workspaces run against customer systems, files, credentials, and approval policies, so evaluation access is provisioned by Duvo after a technical demo or approved pilot. ## How To Request Sandbox Access 1. Visit /contact and request a technical demo or sandbox workspace. 2. Share the workflow category, target systems, approval requirements, and data residency needs. 3. Duvo provisions workspace access, API credentials, MCP access, and sample assignments when the pilot scope is approved. ## First Agent-Ready Steps - Start with the docs hub at https://docs.duvo.ai/ for assignment fundamentals, quick start, building assignments, running assignments, connections, skills, teams, and admin guidance. - Read /developers.md for the developer platform overview. - Install the Duvo CLI with `npm install -g @duvoai/cli` (npm package @duvoai/cli, binary `duvo`) and run `duvo --help`. See /cli.md for the full guide. - Read /auth.md before granting agents access to a workspace. - Read /api-reference.md or /openapi.json before building an API client. - Read /mcp.md before connecting hosted or custom MCP tools. - Read /webhooks.md before subscribing to run, case, approval, or assignment events. - Read /pricing.md for public pricing and sandbox-access constraints. --- # Duvo Pricing Duvo pricing is scoped by workspace, process, systems, support model, and governance requirements. Public no-signup pricing is not currently listed on duvo.ai. ## Evaluation Access - Public no-signup sandbox: not currently available. - Evaluation workspace: provisioned by Duvo after a technical demo or approved pilot. - API, MCP, webhook, and sandbox access: provisioned for approved customer or pilot workspaces. ## What A Pilot Usually Scopes - Workflow category and expected volume. - Target systems, files, portals, and connections. - Required approval gates and audit evidence. - Data residency, SSO, RBAC, and security requirements. - Success criteria and production rollout path. Contact Duvo at /contact for pricing, sandbox access, and enterprise evaluation. --- # Duvo Compare Built for retail and CPG teams whose work spans SAP, portals, files, and approvals, not just clean SaaS APIs. Duvo maps the real process, then runs governed agents against it. ## Positioning Boundary Do not claim that competitors lack agents, governance, orchestration, audit trails, or human control. ServiceNow, Salesforce Agentforce, UiPath, Automation Anywhere, Workato, Power Automate, Copilot Studio, n8n, Zapier, Make, LangGraph, Temporal, CrewAI, and OpenAI Agents SDK all cover meaningful parts of the modern agentic automation stack. Duvo wins when the buyer needs one operating layer that starts with real process truth, improves the process before automation, governs execution against the approved map, applies retail and CPG domain skills, and runs the outcome across the existing retail/FMCG/CPG stack. Duvo is built for retail, FMCG, and CPG teams whose work crosses SAP ECC, SAP S/4HANA, SAP Business One, ERP, supplier portals, spreadsheets, email, files, approvals, and exceptions. ## Where Duvo Wins - Operating truth: Duvo Clarity captures how supplier, inventory, finance, category, and compliance work actually happens across systems, screens, documents, interviews, handoffs, and exceptions. - A better process: Duvo does not treat the current path as sacred. It identifies what should be redesigned, automated, approved, escalated, or left with a human before agents are put in production. - Compliance by map: The approved process map becomes the control surface for execution: allowed actions, exception catalogues, evidence requirements, approval gates, and operating rules. - Real-stack execution: Duvo Automation can run the improved process across ERP, SAP, supplier portals, email, spreadsheets, files, browser and desktop systems, humans, and specialist agents. ## Process truth before automation They need to know how the work really runs, where the process is broken, which exceptions matter, which decisions need approval, and what evidence must exist before the case can close. Duvo starts there, then turns that approved model into agent work. - Capture: Map the live process. Capture recordings, interviews, documents, system evidence, handoffs, rework loops, and exception paths instead of relying on the written SOP alone. - Improve: Decide the operating model. Separate routine agent work from redesign, review, escalation, policy gates, commercial judgment, and human ownership before shipping automation. - Run: Run cases against the map. Turn the process into assignments, queues, tools, evidence checks, approval states, specialist handoffs, writebacks, and audit history. ## Mapped-Process Compliance - Approved process map: The target workflow, exception catalogue, system responsibilities, decision thresholds, and human gates become the reference model for execution. - Evidence requirements: Runs can collect files, messages, system checks, screenshots, exported tables, approvals, and decision notes before closing a case. - Exception ownership: When work diverges from the map, Duvo routes to the right queue, specialist skill, or reviewer with context attached. - Audit-ready outcome: The completed case keeps the result, writeback, tool calls, source files, approvals, and decision trail together. ## Enterprise Agent Control For Real Operations - Case queues and ownership - Human approvals and gates - SAP, ERP, and system access - Files and evidence packs - Schedules and webhooks - Specialist agent handoffs - Policy and operating rules - Replayable run history ## Retail and CPG tools and skills The platform is designed for the kinds of work category, procurement, supply chain, finance, and compliance teams actually run: replenishment, inventory health, input-cost shifts, supplier discovery, label checks, competitive moves, and SAP-backed procurement data. - Automatic ordering: Forecast-to-order workflows with validation gates for forecast quality, inventory position, supplier constraints, budget, logistics, business rules, and external signals. - Inventory optimization: Network inventory analysis for overstock, understock, stockout risk, dead stock, service levels, days of supply, and prioritized action plans. - Input cost monitoring: Commodity and raw-material monitoring for dairy, bakery, produce, proteins, beverages, and household categories, tied back to supplier prices and contracts. - Grocery should-cost: Bottom-up cost models for dairy, bakery, pastry, snacks, confectionery, and frozen products using materials, conversion, packaging, logistics, and supplier margin. - Supplier discovery: Scouting for emerging grocery and CPG suppliers using trend signals, brand research, assortment fit, novelty, and distribution readiness. - Regulatory compliance: Food label, allergen, nutrition, additive, market-entry, and health-claim checks across EU, US, UK, Switzerland, Canada, Australia/New Zealand, and GCC rules. - Competitor intelligence: Retail competitor monitoring across promotions, assortment, private label, launches, digital activity, and market signals. - Case handover: Escalations to category managers, buyers, finance reviewers, compliance owners, and supply-chain specialists with the source evidence attached. ## System reach When a process depends on SAP, on-premise SAP, SAP S/4HANA, legacy ERP, files, supplier websites, private network tools, or Windows desktops, Duvo can combine connectors, browser automation, remote desktop control, SSH/VPN paths, and human approval gates under one governed case. - SAP and ERP: On-premise SAP ECC, SAP S/4HANA, SAP Business One, NetSuite, Oracle Fusion, Microsoft Dynamics 365, Business Central, and other ERP-backed operating systems. - On-premise and private systems: Customer-network paths such as VPN-backed browser access, SSH jump hosts, RDP, and desktop automation for controlled work in private or legacy environments. - SAP data and UI paths: Use SAP table, CDS view, RFC/BAPI, or Fiori browser workflows when those are the access routes the customer provisions. - Data, files, and collaboration: Snowflake, BigQuery, Databricks, Excel, Google Sheets, SharePoint, OneDrive, Google Drive, Outlook, Gmail, Slack, and Teams. ## Vendor And Category Comparison Matrix ### Broad enterprise platforms: ServiceNow, Salesforce Agentforce - What they already do: ServiceNow and Agentforce have agents, governance, data fabrics, control-plane concepts, and broad enterprise platform reach. - Duvo edge: The center of gravity is the operating process, not the platform boundary. - Copy angle: ServiceNow and Agentforce govern agents inside broad enterprise platforms. Duvo governs whether supplier, inventory, category, finance, and compliance work follows the mapped retail or CPG process across SAP, portals, files, emails, spreadsheets, and human decisions. ### RPA and agentic automation suites: UiPath, Automation Anywhere - What they already do: UiPath and Automation Anywhere can orchestrate bots, agents, APIs, process discovery, governance, and automation execution. - Duvo edge: The starting point is the approved operating model, not the automation object. - Copy angle: Agentic automation platforms industrialize bots, agents, and systems. Duvo starts with the live process, the better future process, the domain checks, the approval thresholds, and the evidence an operator needs before the work can close. ### Integration and agent builders: Workato, Power Automate, Copilot Studio - What they already do: These platforms can connect apps, build agents, manage actions, govern flows, and orchestrate work across many systems. - Duvo edge: The deliverable is a closed operational case, not just a successful action chain. - Copy angle: Integration platforms move actions across apps. Duvo carries the case through missing data, approvals, SAP or ERP steps, supplier evidence, human review, writeback, and audit-ready closure. ### Workflow builders: n8n, Zapier, Make - What they already do: Workflow builders can assemble automations, support AI agents, integrate tools, and include human-in-the-loop patterns. - Duvo edge: The buyer gets a governed operating layer, not only a place to assemble flows. - Copy angle: Workflow builders help teams assemble automations. Duvo gives operations leaders reusable process maps, retail and CPG skills, case queues, evidence standards, approval gates, and outcome ownership. ### Agent and workflow frameworks: LangGraph, Temporal, CrewAI, OpenAI Agents SDK - What they already do: Frameworks can provide durable execution, persistence, tracing, tools, handoffs, flows, evals, and human-in-the-loop primitives. - Duvo edge: Duvo is the business operating product around the agent substrate. - Copy angle: Frameworks help developers build reliable agents. Duvo helps enterprises decide what those agents should do, give them approved system access, route their work through operators, and prove the retail or CPG process was followed. ## Canonical URLs - Human page: https://www.duvo.ai/compare - Markdown page: https://www.duvo.ai/compare.md - Contact: https://www.duvo.ai/contact --- ## Source Files - [Homepage markdown](https://www.duvo.ai/index.md) - [Developer platform](https://www.duvo.ai/developers.md) - [API reference](https://www.duvo.ai/api-reference.md) - [Authentication](https://www.duvo.ai/auth.md) - [Webhooks](https://www.duvo.ai/webhooks.md) - [MCP](https://www.duvo.ai/mcp.md) - [CLI](https://www.duvo.ai/cli.md) - [Getting started](https://www.duvo.ai/getting-started.md) - [Pricing](https://www.duvo.ai/pricing.md) - [Compare](https://www.duvo.ai/compare.md) - [Navigation index](https://www.duvo.ai/llms.txt)