Developer Tools
MCP servers, CLIs, SDKs, APIs, and open source tooling for building on agentic AI platforms.
Building on agentic AI platforms requires the right developer tools—from MCP servers and CLIs to SDKs, APIs, and integration frameworks. Explore open source tooling, integration guides, and developer resources for building, extending, and connecting AI-powered applications.
920 articles in this category
Bounded vs Open-Ended AI Engagements: The Difference
An engagement is bounded when the thing being integrated is finished. Everything else — ceilings, change control, weekly reporting — is compensation for a boundary that was never there.
The $7.2M Abandoned AI Initiative: What Goes Wrong
The average abandoned enterprise AI initiative has $7.2M sunk into it, and 88% of pilots never reach production at all. The failure is rarely the model — it's the eighteen months spent building a foundation.
How to Write a Statement of Work for AI Infrastructure
Most AI statements of work define done as a feature list, which makes acceptance a negotiation. Define it as a held-out evaluation set with a passing threshold, and settle source-code rights in the SOW itself.
Legal AI's Next Crisis Is Trust, Not Intelligence
Legal AI agents are drafting motions using static API keys tied to shared service accounts, with no verified identity and no per-action audit trail. The LexisNexis breach confirmed in March 2026 showed what one over-privileged machine identity costs — and the profession's own attribution standards were never written for a caller that is not a person.
Who Should Build Your AI Platform in 2026?
Accenture booked $11.5B in advanced-AI work; OpenAI capitalized a deployment company above $4B; Anthropic's services JV is reported above $1.5B. Three kinds of partner, and what each one structurally cannot give you.
UK Sovereign AI: Real Procurement, But the IP Still Leaves
The UK's £500m Sovereign AI Unit is the most concrete sovereign-AI programme any major government has run — and its own contract terms let suppliers keep all the IP while government retains usage rights only. Meanwhile £1.41bn of 2026 UK public-sector AI procurement still flows mostly to Microsoft and Palantir.
79% of Enterprises Overran Their AI Budget. Here's Why.
79% of enterprises hit AI cost overruns in the past year and 80-85% missed infrastructure forecasts by more than 25%. The driver isn't model licensing or compute — it's the foundation nobody counted.
The Framework War Is About Who Owns the Agent Runtime
Within nine days in spring 2026, Microsoft collapsed Semantic Kernel and AutoGen into a single agent runtime and Intel put 32GB of VRAM in a $949 card. Those two events point in opposite directions, and the choice between them is not about features — it is about who owns the runtime your agents execute on.
Time and Materials Is an Admission, Not a Pricing Model
FAR permits time-and-materials only when it is impossible to estimate the work, and says outright that T&M gives the contractor no incentive to control cost. Both sentences describe a vendor starting from zero.
The Inference Era: Why AI Pricing Has to Move Past Per-Seat
Hyperscaler capex is heading for $660-690 billion in 2026 and the money is moving from training to inference — yet enterprises still buy AI by headcount. The per-seat sticker price is also not the per-seat price: Microsoft 365 Copilot's $30 add-on is $69 to $90 a seat once the required base licenses are counted.
Healthcare AI Is Consolidating Into an Operating System
Scheduling, triage, documentation, and billing are converging from separate AI vendors into one platform. McKinsey calls it a modular architecture — the question health systems should ask is who owns the layer everything else plugs into.
Banks Are Building AI Workforces on Infrastructure They Rent
Banks are deploying agents for KYC, compliance, and fraud detection — but Capgemini finds only 10% run them at scale, and most run on infrastructure the bank does not own. Why the second fact explains the first.
The Model Is a Commodity. The Operating System Is the Moat.
Alibaba's Qwen crossed 3 billion downloads and open weights now match frontier performance at a fraction of the cost, which means the model is no longer where advantage lives. The durable layer is the operating system around it — and we shipped 40 production releases into ours in a single week to make the point concrete.
FERPA Governs Data, Not Which Model Reasons About It
Alibaba's Qwen crossed 3 billion downloads to become the most-downloaded open model family, and open weights now sit under products of every origin. FERPA regulates who may access an education record — it says nothing about which model processes it or where inference runs, and that gap has to be closed in the contract.
The IMO-Perfect Model's Open Sibling: Reasoning You Can Host
RedNote's dots-note-3.0 scored a certified 42/42 at the 2026 IMO. Its open-weight sibling, dots3-note-preview, shipped under Apache 2.0 on August 14 — 280B parameters, 16B active, 512K context. What that separation actually means for owning frontier reasoning.
Beyond LLMs: What Reasoning Limits Mean for Clinical AI
A widely-shared DeepMind position paper argues LLMs cannot make the abductive leap that produces new scientific theories. It is a narrower claim than the headlines suggest, and it is not the reason clinical AI fails today — but it does explain why a health system should build for model replacement rather than model selection.
K-12 AI Adoption Is Outpacing Its Safety Infrastructure
K-12 is adopting AI faster than any other education segment and has the least infrastructure to govern it. What district-grade AI safety actually requires — and why the model-ownership question decides most of it.
AI Governance: Enterprise Software's Fastest-Growing Category
Vals AI raised a $40M Series A at a $400M valuation for a product that validates other companies' AI rather than building models. That is a category forming around a measurement gap — and the reason the gap exists is that most enterprises are trying to govern systems they cannot inspect.
Sovereign or Supervised: Government AI Architecture
In one week of August 2026 the EU moved to restrict foreign cloud providers from sensitive public-sector workloads, and researchers documented autonomous AI agents breaching 85 Taiwanese government accounts. Read together, the two events make the same argument: for a government agency, where AI runs is a security architecture decision rather than a procurement preference.
Kenya's Draft AI Policy Spreads Liability Across the Chain
Kenya's draft AI policy proposes allocating liability across the entire chain — developers, deployers, operators, vendors, and users. The US is still debating timelines. The interesting question is not who moved first but why a jurisdiction without entrenched technology lobbies produced a cleaner rule, and what full-chain liability means for anyone deploying AI on someone else's infrastructure.
Google Cloud's 20 Questions Before Deploying AI Agents
Google Cloud published a governance checklist for organizations deploying production AI agents rather than another capability announcement. That inversion is the signal worth reading: the constraint on agentic deployment has moved from what models can do to what organizations can defend. Several of the questions cannot be answered at all on infrastructure you do not control.
The Database Layer Went Agentic: PGBot and Postgres
PGBot is a free, open-source Go tool that gives AI agents native PostgreSQL intelligence — schema reasoning and query optimization without a human translating between the model and the database. It marks a shift worth understanding: the data layer is becoming something agents reason about directly, which makes who controls that layer the deciding question.
Code Mode: One Prompt to a Running Next.js App You Own
Code Mode takes a prompt and returns a scaffolded Next.js app with components installed and the dev server running. Agent Skills make the playbooks behind it reusable across agents. The interesting part is not the speed — it is that the output is a codebase in your repository rather than an app inside someone else's platform.
When Compliance AI Hallucinates, Who Audits the Filing?
A 125-year-old law firm was ordered to explain AI-hallucinated citations in a court brief. The same class of tool now drafts SEC and FINRA filings, where the reviewer is an examiner rather than a judge. The difference between a sanction and a clean examination is whether you can reconstruct what the model saw — which is an infrastructure property, not a model one.