Foundry
An enterprise-grade AI model hosting and development platform offered by Microsoft Azure. It supports the deployment of and API access to a variety of large models, and is designed for enterprise compliance and data residency requirements.
Core Facts
Timeline (last 90 days)
Palantir's core products Gotham and Foundry are not out-of-the-box standardized software, but platforms that require engineers to intervene deeply in customers' business processes for customized configuration.
/auto-mode-setup covers Bedrock, Vertex, Foundry, Claude apps gateway, and sessions with telemetry disabled.
Fab (wafer manufacturing) refers to companies that own wafer manufacturing lines, responsible for turning chip designs into physical wafers, a capital-intensive segment where a single 12-inch production line often requires billions of dollars of investment; Fabless refers to fabless design companies that only do chip design and outsource manufacturing to foundries; Foundry refers to the pure-play foundry model (such as TSMC); IDM refers to the integrated design-and-manufacture model (such as Silan).
Palantir's best-known products include Gotham, aimed at government and defense, and Foundry, aimed at commercial clients.
Microsoft released Decision-1 in Foundry, a post-trained Qwen3.5-9B model designed specifically for single-shot decision scoring, priced at $0.042 per million input tokens, with output free.
The endpoint path for Microsoft-Decision-1 on Foundry is <your-foundry-resource-endpoint>/providers/microsoft/v1/systemone, and the model field in the request is filled with the deployment name rather than the model string.
The officially permitted uses of subscription-included API credits include: claude -p or the Agent SDK run directly with an API key, the Claude API and Console Playground; interactive Claude Code (terminal, IDE, desktop, web) as well as Bedrock, Vertex AI and Foundry are not available.
Sample code should use the resource endpoint https://your-resource.services.ai.azure.com rather than the Foundry project endpoint ending in /api/projects/...
Microsoft-Decision-1 has gone live on the Microsoft Foundry and OpenRouter platforms
Opus 4.7+ 和 Fable 模型现在在 Bedrock、Vertex、Foundry 以及 Claude apps 网关上默认使用100万(1M)上下文窗口,不再需要 [1m] 后缀
16 more timeline events
All Facts (20)
Palantir was founded in 2003 by Peter Thiel and others and is a data analytics company.
80%VerifiedPalantir's two core product lines are the Gotham platform for intelligence analysis and the Foundry platform for commercial data integration.
65%UnverifiedPalantir's core products include Palantir Gotham and Foundry, which are closed-source commercial software primarily serving governments and large enterprises.
60%UnverifiedPalantir primarily served the US government and intelligence agencies in its early days, with products including Gotham and Foundry.
60%UnverifiedPalantir's core products Gotham and Foundry are not out-of-the-box standardized software, but platforms that require engineers to intervene deeply in customers' business processes for customized configuration.
50%Unverified/auto-mode-setup covers Bedrock, Vertex, Foundry, Claude apps gateway, and sessions with telemetry disabled.
50%UnverifiedFab (wafer manufacturing) refers to companies that own wafer manufacturing lines, responsible for turning chip designs into physical wafers, a capital-intensive segment where a single 12-inch production line often requires billions of dollars of investment; Fabless refers to fabless design companies that only do chip design and outsource manufacturing to foundries; Foundry refers to the pure-play foundry model (such as TSMC); IDM refers to the integrated design-and-manufacture model (such as Silan).
50%UnverifiedPalantir's best-known products include Gotham, aimed at government and defense, and Foundry, aimed at commercial clients.
50%UnverifiedThe endpoint path for Microsoft-Decision-1 on Foundry is <your-foundry-resource-endpoint>/providers/microsoft/v1/systemone, and the model field in the request is filled with the deployment name rather than the model string.
50%UnverifiedThe officially permitted uses of subscription-included API credits include: claude -p or the Agent SDK run directly with an API key, the Claude API and Console Playground; interactive Claude Code (terminal, IDE, desktop, web) as well as Bedrock, Vertex AI and Foundry are not available.
50%UnverifiedSample code should use the resource endpoint https://your-resource.services.ai.azure.com rather than the Foundry project endpoint ending in /api/projects/...
50%UnverifiedPalantir's core product Foundry is designed around the concept of Ontology, defining internal enterprise entities and their relationships as a unified structured semantic graph.
50%UnverifiedIn a Foundry Canvas deployment, azure.yaml is responsible for provisioning the Foundry project and model deployment, while toolbox.yaml serves as a placeholder configuration for connecting the toolbox.
50%UnverifiedFoundry Hosted Agent provides two independent isolation control mechanisms: user isolation and session isolation.
50%UnverifiedFoundry session isolation identifies a VM-level isolated sandbox within a managed session, in which code, files, and working state are all persisted.
50%UnverifiedClaude Code's AGENTS.md compatibility rules can be modified in Config's Project Instructions, but currently do not cover hosted environments such as Bedrock, Vertex, and Foundry.
50%UnverifiedThe server-side classifier default switch applies to Claude API and enterprise users, as well as deployments running on Amazon Bedrock, Google Vertex, Foundry, and various gateways.
50%UnverifiedThe maxEffortLevel configuration covers all API providers, including Bedrock, Vertex, and Foundry.
50%UnverifiedBedrock, Vertex, Foundry, and other sessions now explicitly inform Claude when an MCP server connection fails, rather than letting it mistakenly assume the tools do not exist
50%UnverifiedFixed credential and authentication header leakage issues on multiple platforms including Bedrock, Vertex, and Foundry, including a problem where residual Anthropic API keys were being sent alongside Foundry subscription key mode requests.
50%