# Duale AI > Duale AI is a production runtime that keeps Python agent contracts stable while models, providers, and routing policy evolve. Duale AI is a production runtime for durable AI agents, supporting Python as the developer path for contracts, deadlines, typed results, and provider choices. Teams keep their own model provider accounts; Duale AI routes to enabled providers through configurable policy without rewriting dependent agents. Task deadlines are absolute dates defaulting to 30 minutes ahead, configurable to hours or days; after a deadline passes the task fails with no partial result. Pricing separates managed runtime orchestration measured in AIOp from model inference, which is billed directly by the customer's chosen providers. Managed application data is hosted by Hetzner in Germany; Duale AI is a French company, and customers own policy choices, provider contracts, and retention requirements. Governance controls give security, legal, audit, and platform teams a common view of agent data movement, provider choices, and review paths. ## Pages ### Production runtime - [Production runtime for durable AI agents](https://dev.duale.ai/index.md): The Duale AI production runtime keeps Python agent contracts stable while models, providers, and routing policy evolve. ### Solutions - [One runtime path for production agents by operating role](https://dev.duale.ai/en/solutions.md): A shared production layer for durable Python agents, organized by operating role, provides stable contracts, routing policy, deadlines, and recovery behavior. - [Turn AI pilots into a production portfolio](https://dev.duale.ai/en/solutions/business.md): Duale AI's shared production runtime lets organizations scale agents with lifecycle control, cost visibility, and provider choice. - [Govern AI agents while they move to production](https://dev.duale.ai/en/solutions/governance.md): Duale AI's shared operating model gives security, legal, audit, and platform teams a common view of agent data movement, provider choices, and review paths. ### Product - [Runtime security and pricing for production agents](https://dev.duale.ai/en/product.md): Duale AI's production agent runtime is secured through approved model routing and priced for work returning typed results. - [Security controls for production agents](https://dev.duale.ai/en/product/security.md): Duale AI's managed service provides security controls, privacy boundaries, EU hosting details, and deployment-specific review paths for evaluating teams. - [Pricing for reliable agent orchestration](https://dev.duale.ai/en/product/pricing.md): Duale AI pricing separates managed runtime orchestration from model inference, which your chosen providers bill directly. ### Documentation - [Technical documentation for developers](https://dev.duale.ai/en/docs.md) (General Availability): The public integration contract for developers defines the Python SDK as the main application interface and four platform behaviors outside the task loop. - [Release stages: access, support, and change policy](https://dev.duale.ai/en/docs/release-stages.md) (General Availability): Duale AI's release stage policy defines access, production qualification, support, and change policy for each stage from Private Preview through Retired. - [Connect an MCP client to the documentation server](https://dev.duale.ai/en/docs/mcp.md) (Public Preview): The Duale AI MCP server lets clients search documentation and fetch complete pages with canonical URLs without an account or API key. - [Python SDK for bounded agent work with typed results](https://dev.duale.ai/en/docs/sdk.md) (Public Preview): The Duale AI Python SDK submits bounded agent work from Python and returns a validated typed result. - [SDK task lifecycle, routing, and streaming](https://dev.duale.ai/en/docs/sdk/concepts.md) (Public Preview): The Duale AI SDK submits bounded tasks to a hosted runtime that selects model routes and streams content deltas with a terminal result. - [Declare and serve Python tools with the SDK](https://dev.duale.ai/en/docs/sdk/tools.md) (Public Preview): The Duale AI SDK lets you declare Python functions with @tool and publish them via sdk.serve\(\) so a task can call them. - [Attach files to a task using the SDK](https://dev.duale.ai/en/docs/sdk/attachments.md) (Public Preview): The Duale AI SDK attachment workflow prepares, uploads, and submits files to a task, resolving them into a conventional Library path. - [Manage Libraries and documents with the SDK](https://dev.duale.ai/en/docs/sdk/manage-libraries.md) (Public Preview): The Duale AI SDK manages a persistent Library and its documents, covering upload, ingestion polling, state inspection, and deletion. - [SDK exception hierarchy, problem details, and retry behavior](https://dev.duale.ai/en/docs/sdk/errors.md) (Public Preview): The Duale AI SDK exception hierarchy and RFC 9457 problem details envelope define how to classify, retry, and report task and Library failures. - [Application-facing SDK API reference](https://dev.duale.ai/en/docs/sdk/reference.md) (Public Preview): The Duale AI SDK application API defines task submission, conversation continuation, LibrariesClient, attachment, tool, and configuration interfaces. - [Understand the agent harness contract](https://dev.duale.ai/en/docs/agent-harness.md) (Public Preview): The agent harness coordinates model calls, tools, and delegated executors between one task submission and one terminal outcome. - [Understand nested executors and deadlines](https://dev.duale.ai/en/docs/agent-harness/executors.md) (Public Preview): One SDK submission divides work across an executor tree and returns one terminal outcome within a single task deadline. - [Keep context across long-running work](https://dev.duale.ai/en/docs/agent-harness/context-continuity.md) (Public Preview): Duale AI keeps long-running work coherent by separating stored task history from the active context sent to a model. - [The documents your agents can read](https://dev.duale.ai/en/docs/libraries.md) (Public Preview): A Library holds your organization's documents so a granted agent can read them during a task without writing retrieval code. - [How an agent finds an answer in your documents](https://dev.duale.ai/en/docs/libraries/what-agents-can-do.md) (Public Preview): An agent answers document questions by choosing between semantic search and exact string match, so corpus shape determines which questions get usable answers. - [Design your Libraries by reader group and agent reach](https://dev.duale.ai/en/docs/libraries/design.md) (Public Preview): Library grouping in Duale AI follows reader groups, because a grant covers a whole Library and no per-document grants exist. - [Grant and review access to a Library](https://dev.duale.ai/en/docs/libraries/access.md) (Public Preview): Resource grants control which subjects reach a Library through five scoped actions, with refusals overriding every allow. - [How the platform protects and removes your documents](https://dev.duale.ai/en/docs/libraries/security.md) (Public Preview): Duale AI encrypts Library documents at rest, verifies integrity before agent access, and removes them on a 30-day reversible schedule. - [Library limits for sizing files, batches, and storage plans](https://dev.duale.ai/en/docs/libraries/limits.md) (Public Preview): Library limits define file size, format support, API ceilings, time windows, and metering for sizing files, batches, and storage plans. - [Get better answers from a Library](https://dev.duale.ai/en/docs/libraries/best-practices.md) (Public Preview): Document preparation and management practices that improve or protect the material an agent reads from a Library. - [Architecture patterns for Libraries](https://dev.duale.ai/en/docs/libraries/patterns.md) (Public Preview): Methods for ensuring document extraction fidelity in Library cover manual extraction of critical documents and reporting extraction failures. - [Web search: delegated, source-led research for agent tasks](https://dev.duale.ai/en/docs/web.md) (Public Preview): Web research in Duale AI delegates a bounded question to a dedicated executor that searches, reads, and synthesizes public pages. - [How Web search works: delegated research and evidence](https://dev.duale.ai/en/docs/web/how-web-search-works.md) (Public Preview): Duale AI Web search delegates a bounded question to a dedicated executor that searches, reads, and returns a compact evidence package. - [Request Web research with source links](https://dev.duale.ai/en/docs/web/request-web-research.md) (Public Preview): Requesting Web research with source-linked claims lets your application reject incomplete results before a person reviews the sources. - [Sources, citations, and factuality](https://dev.duale.ai/en/docs/web/sources-and-factuality.md) (Public Preview): Web research factuality distinguishes discovery results, sources, citations, grounded claims, and fact-checks, and defines verification of material claims against evidence. - [Page access for the Web research executor](https://dev.duale.ai/en/docs/web/page-access.md) (Public Preview): Page access lets the Web research executor read a public HTTP or HTTPS URL, rendering JavaScript and returning extracted Markdown, title, and final URL. - [Troubleshoot Web search when results are missing or blocked](https://dev.duale.ai/en/docs/web/troubleshooting.md) (Public Preview): Duale AI Web search troubleshooting covers missing results, blocked pages, absent sources, and task deadline failures. - [Intended purpose, excluded uses, and behavioral constraints](https://dev.duale.ai/en/docs/scope-and-limits.md) (General Availability): Duale AI's Platform states its intended purpose, excluded uses under Regulation \(EU\) 2024/1689 Annex III, and behavioral constraints. - [Secure your use of the platform](https://dev.duale.ai/en/docs/security.md) (Public Preview): Duale AI security controls are organized around six moments in a task lifecycle, defining where platform protections end and your integration's responsibility begins. - [Divide security responsibility](https://dev.duale.ai/en/docs/security/shared-responsibility.md) (Public Preview): The shared responsibility model divides platform operations from agent control, data ownership, and decision authority between Duale AI and the customer. - [Protect data across every boundary](https://dev.duale.ai/en/docs/security/data-protection.md) (Public Preview): Duale AI's data protection model defines encryption at rest, SDK-to-platform transport protection, and the external systems a task can reach. - [Know what the platform defends](https://dev.duale.ai/en/docs/security/threat-model.md) (Public Preview): The Duale AI threat model states where untrusted text enters the model context, what the platform bounds, and what your integration must add. - [Build an agent that passes review](https://dev.duale.ai/en/docs/security/secure-integration.md) (Public Preview): Security controls you must implement in your own code close gaps the platform does not close for you. - [Know what the platform records](https://dev.duale.ai/en/docs/security/evidence-and-audit.md) (Public Preview): The Duale AI audit log records platform events and authorization decisions but omits document reads, policy changes, and tool results you must record yourself. - [Content marking for terminal text answers](https://dev.duale.ai/en/docs/content-marking.md) (Public Preview): Duale AI marks terminal plain-text answers with machine-readable C2PA provenance to support the text-marking requirement in Article 50\(2\) of Regulation \(EU\) 2024/1689. - [Understand model routing and the stable application contract](https://dev.duale.ai/en/docs/model-routing.md) (Public Preview): Model routing keeps one application contract across an approved set of external model routes through six control layers. - [Routing contract: hard boundaries and soft preferences](https://dev.duale.ai/en/docs/model-routing/contract-and-limits.md) (Public Preview): The routing contract separates hard application boundaries from soft preferences that vary as pool, context, and operating conditions change. - [Provider capability and compatibility](https://dev.duale.ai/en/docs/model-routing/provider-capabilities.md) (Public Preview): Route fields and provider limits affect structured output, context caching, image transport, reasoning, and token counting in Duale AI model routing. - [Integrate model routing in an application](https://dev.duale.ai/en/docs/model-routing/application-integration.md) (Public Preview): The Duale AI SDK contract carries business intent and routing preferences to every qualified route that supports your application's functions. - [Plan the deployment and model portfolio](https://dev.duale.ai/en/docs/model-routing/plan-deployment.md) (Public Preview): Deployment boundary planning maps hard constraints across tenants, regions, and networks before selecting a model pool shape and portfolio. - [Model routing architecture patterns](https://dev.duale.ai/en/docs/model-routing/architecture-patterns.md) (Public Preview): Model routing architecture patterns show how to structure model pools based on hard boundaries and failure outcomes rather than preferred provider names. - [Control model changes across the production lifecycle](https://dev.duale.ai/en/docs/model-routing/control-model-changes.md) (Public Preview): The model change control process defines evaluation, enablement, observation, rollback, emergency cutoff, and retirement of production model routes. - [Operate model routing and manage risk](https://dev.duale.ai/en/docs/model-routing/operations-and-risk.md) (Public Preview): Operating model routing as a business-process service requires monitoring health across business, task, and control boundaries. - [Ownership and approvals for routing decisions](https://dev.duale.ai/en/docs/model-routing/ownership-and-approvals.md) (Public Preview): Routing decision ownership assigns responsibility across seven actions and ten lifecycle events from proposal through stop. - [Divide agent ownership and access](https://dev.duale.ai/en/docs/agents-and-access.md) (Public Preview): An agent is the unit your organization gives to one team, carrying its own identifier, tokens, tools, and records. ### Legal - [Legal terms, privacy policy, and subprocessors](https://dev.duale.ai/en/legal.md): Duale AI legal and privacy documents provide the governing terms, policies, and safeguards for platform use and data handling. - [Terms of use for AI agent orchestration](https://dev.duale.ai/en/legal/cgu.md): Duale AI's Terms of Use set the rules for using its AI agent orchestration Platform, covering data processing, model routing, and prohibited uses. - [Terms of sale for the prepaid credit platform](https://dev.duale.ai/en/legal/cgv.md): These terms define the prepaid credit model for Duale AI's SaaS AI agent orchestration platform, covering pricing, wallet mechanics, termination, and liability. - [Privacy policy for personal data and AI services](https://dev.duale.ai/en/legal/privacy.md): Duale AI's privacy policy describes how the company collects, uses, and protects personal data across its website, SaaS platform, and communications. - [Subprocessors and data transfer safeguards](https://dev.duale.ai/en/legal/subprocessors.md): Duale AI lists the subprocessors that process personal data for its services and the transfer safeguards applied to each. ## Full Markdown Inventory - [Markdown sitemap](https://dev.duale.ai/sitemap.md): Exhaustive Markdown mirror inventory for all generated pages and languages.