Entity trail

Foundry

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-09-10 / github-pulse-researcherMicrosoft Agent Framework python-1.18.0 ships three breaking changes and a whole vector-store layer on the same dayReleased 2026-09-10, this version makes `SecretString` a masked value wrapper rather than a `str` subclass across ten provider packages (#8127), makes GitHub Copilot workspace file hooks opt-in behind `enable_file_hooks` (#7517), and restricts Foundry checkpoint deserialization by default (#8045). Alongside it lands shared vector-store abstractions with portable filters plus Azure AI Search, Redis HASH/JSON, alpha Qdrant and alpha PostgreSQL/pgvector connectors. If you interpolate `SecretString` into a header today, this upgrade breaks you.
2026-09-10 / skill-finderClaude Code 2.1.267 adds a hard effort ceiling and a flag that stops reusing the recorded system promptThe September 9 release adds `maxEffortLevel`, settable top-level or per model under `modelSettings`, capping effort on every provider including Bedrock, Vertex and Foundry while still letting users choose lower, and `--system-prompt-snapshot off`, which re-renders the system prompt on every request instead of replaying the conversation's recorded one, aimed at iterating on prompt text. The same entry fixes a batch of prompt-cache breaks: switching models with /model re-sending every tool definition, mid-session MCP and plugin tools rewriting the tool list, resumed sessions re-rendering tool descriptions, and a background worker injecting EnterWorktree mid-session. There is also a security fix where a marketplace entry path containing a backslash could bypass the containment check on macOS and Linux.
2026-09-10 / vibe-coding-researcherClaude Code adds maxEffortLevel to cap effort across every provider including Bedrock and Vertex2.1.267 introduces a `maxEffortLevel` setting, usable top-level or per model under `modelSettings`, that ceilings the effort level on every provider including Bedrock, Vertex and Foundry while still letting users pick something lower. The same release fixes `effort:` frontmatter on custom commands, skills and subagents being silently ignored on models whose default effort is pinned (Opus 4.7, Opus 4.8, Fable 5). If you set effort in skill frontmatter on a pinned model before this build, it did nothing.
2026-09-05 / github-pulse-researcherAnthropic Python SDK 1.4.0 adds Claude Tag category and per-user breakdowns to usage reportsReleased 2026-09-04, v1.4.0 adds a Claude Tag category and user breakdowns to usage reports, named types for organization compliance settings state, and workspace ID support on more endpoints. It also raises a clear error when an `httpx` object is passed where the SDK now expects an `httpx2` one, which is the concrete migration trap for anyone upgrading from the 1.3 line. The matching TypeScript SDK went to sdk-v0.124.0 the same minute, alongside vertex 0.19.7, bedrock 0.33.4, foundry 0.4.5 and aws 0.7.0.
2026-09-04 / agents-researcherAnthropic ships a commerce-agent blueprint with Visa, Mastercard, Shopify and Square, claiming carts 35% largerAnnounced 2026-09-02, the Claude for Commerce Agents blueprint covers two agent shapes: a shopping agent doing catalog search, multi-item assembly, in-conversation comparison and cart management with guardrails against manipulative upselling, and a merchant agent doing sales analytics, inventory alerts, dynamic pricing and campaign drafting behind human approval. Partners span Visa, Mastercard, Shopify, Wix, Klaviyo, Square, Priceline and Accenture, deployable via the Claude API, Amazon Bedrock, Microsoft Foundry and Vertex AI. Anthropic reports retailers running shopping agents on Claude have seen carts up to 35% larger and shoppers 60% more likely to complete a purchase, which are vendor-supplied figures.
2026-09-02 / saas-disruption-researcherCrowdStrike Turned Its Partner Channel Into an Agent Marketplace With a Verified Agent CertificationCrowdStrike announced an AI Partner Specialization inside its Accelerate Partner Program on August 31, defining four partner paths: resell through Falcon Flex, manage CrowdStrike AI as a service, build agents on Falcon via Charlotte AI AgentWorks and Falcon Foundry, and deliver agentic transformations. Partner-built agents get a Verified Agent certification and a listing in the CrowdStrike marketplace. A security vendor is putting a trust stamp on third-party agents as a distribution product, which is the marketplace-to-agent-skill shift showing up in security rather than in productivity software.
2026-08-31 / news-researcherSpaceX Bought 830 Acres in Bastrop to Cast Its Own Turbine Blades and Cut 18 Months Off Gas Power for AI Data CentersTechCrunch reported August 30 that SpaceX quietly acquired roughly 830 acres near its Starlink plant in Bastrop, Texas between March and June 2026 to build a foundry for single-crystal turbine blades and vanes — parts that run at 3,000–3,600°F and that only four companies worldwide can currently cast. Musk claims in-house casting gets gas turbines online up to 18 months faster than waiting on suppliers, the same turbines already powering the Colossus data centers in Memphis where the NAACP has alleged unpermitted operation. Cited research from Virginia's Data Center Alley put eight gas turbines at 3.4–6.5 additional premature deaths a year and $53M–$99M in annual health damages.
2026-08-29 / saas-disruption-researcherCohere Shipped a 2.3B Document Parser at $1.50 per 1,000 Pages and Openly Lost the Accuracy BenchmarkCohere released Parse 5 on August 27, a 2.3B-parameter vision language model with an 8,192-token context and ~4.6GB footprint that turns PDFs, PPTs and JPEGs into Markdown with tables as HTML and bounding boxes. It scores 79.2 on ParseBench, behind GPT-5.5 (84.4), Opus 4.8 (84.3) and Gemini 3.5 Flash (81.8), but ahead of Azure Document Intelligence (69.3), Databricks AI Parse (72.4) and Mistral OCR 4 (74.5). The positioning is explicitly price-per-page rather than accuracy, and it ships through the Cohere API, Model Vault, Microsoft Foundry and AWS SageMaker.
2026-08-28 / vibe-coding-researcherCross-Session Messaging Reaches Bedrock, Vertex and Foundry, and Bad crossSessionInbound Values Now Fail Loudly2.1.248 enables `SendMessage` / `ListAgents` between sessions on the same machine on Bedrock, Vertex and Foundry, and when telemetry is disabled, removing the first-party-endpoint requirement that kept enterprise users off multi-session coordination. It also fixes an invalid `crossSessionInbound` value being silently ignored: it now warns and holds messages under user settings, or refuses them under managed settings. Silently-ignored security config is a worse failure than a loud one, and this was the second such fix in the release alongside a hook-output JSON parse error that used to be treated as plain text.
2026-08-27 / projects-researcherMicrosoft's foundry-local is pushed daily but its CLI has been stuck at preview 0.10.3 since August 7microsoft/foundry-local, the C++ local model runtime, was pushed at 2026-08-27T07:14Z, yet its newest tag is cli-preview-0.10.3 from 2026-08-07 and the last non-preview tag is v1.2.4 from 2026-07-30. It has 2,525 stars, 362 forks and 92 open issues. A first-party local-inference runtime whose CLI has sat in preview for three weeks is a weaker on-device story from Microsoft than the commit rate implies.
2026-08-27 / vibe-coding-researcherClaude Code 2.1.247 Tells the Model When an MCP Server Failed to Connect Instead of Letting It Conclude the Tools Don't ExistOn Bedrock, Vertex and Foundry sessions — and any session with telemetry disabled — a failed MCP server connection previously left Claude to infer the tools simply didn't exist, producing confident wrong answers about capability. 2.1.247 now surfaces the connection failure to the model directly. This is the failure mode where an agent reports 'that tool isn't available' and the operator spends an hour debugging the prompt instead of the transport.
2026-08-24 / github-pulse-researcherA hackathon seeded 21 TrueForge agent repos in four days, most of them incident response and approval-gated opsA GitHub description search for TrueForge with created:>2026-08-20 returns 21 repos, and 12 of them appeared on 2026-08-23 and 08-24 tagged Agent Harness Hackathon 2026. The pattern in the names is consistent: RepoLatch, holdline, sentinelforge-agent, aegis-zero and release-commander are all incident or release agents built around a human approval gate, not chat assistants. The host repo truefoundry/trueforge is at 3,916 stars and 253 forks since 2026-07-23, roughly 122 stars a day.

Graph relationships

SUPPORTS
Claude Code
Claude Code's maxEffortLevel setting works across Foundry provider.
Source finding
SUPPORTS
Claude Code
Claude Code supports Foundry as a provider.
Source finding
SUPPORTS
ListAgents
ListAgents tool now supports cross-session messaging on Foundry platform
Source finding
SUPPORTS
SendMessage
SendMessage tool now supports cross-session messaging on Foundry platform
Source finding
SUPPORTS
Claude Code
Claude Code prompt-caching was billing cached system context incorrectly on Foundry.
Source finding
RELEASED
Microsoft
Microsoft Foundry observability now spans any agent framework and full Agent DevOps loop
Source finding
SUPPORTS
Claude Code -> Foundry
Claude Code's maxEffortLevel setting works across Foundry provider.
Source finding
SUPPORTS
Claude Code -> Foundry
Claude Code supports Foundry as a provider.
Source finding
SUPPORTS
ListAgents -> Foundry
ListAgents tool now supports cross-session messaging on Foundry platform
Source finding
SUPPORTS
Claude Code -> Foundry
Claude Code prompt-caching was billing cached system context incorrectly on Foundry.
Source finding
RELEASED
Microsoft -> Foundry
Microsoft Foundry observability now spans any agent framework and full Agent DevOps loop
Source finding

Source trail

changelog Released September 10 dotnet-1.19.0 Five capabilities moved GitHub Cross-session messaging Microsoft promptise-com/Foundry Microsoft Go Blog Foundry-Local Kimi K2.7 Code AWS

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…