#docs

Designing or marketing an API where an autonomous agent could evaluate, integrate, or transact on a developer's behalf — anything that sells or provisions something programmatically.

Planning docs work, an information-architecture overhaul, or deciding whether to publish llms.txt.

Your product ships an MCP server, or "MCP support" appears anywhere in your positioning.

Docs served as markdown and OpenAPI with a quickstart that says "hand this spec to your LLM" — and a purchase flow (quote-token, idempotency keys, consent record) an agent can run without over-buying.

A three-pane API reference with runnable, copy-once code that carries your own test key — the docs are the product tour.

"Engineers who write" as the staffing model, with Dell, NVIDIA, Qualcomm, AMD and Stripe named, sold three ways: a scoped sprint, an embedded documentation engineer, or full ownership.

300+ vetted engineer-writers and a stated 100+ devtool clients, Docker, JetBrains and Redpanda among the logos on the wall.

50+ engineering-first clients including Cloudinary, Auth0, Netlify, Appwrite and Sourcegraph, drawn from a stated 1,500-strong technical creator network.

In-house engineer-writers rather than a freelancer marketplace, 40+ infrastructure and devtool logos including Travis CI and env0, and a stated seven-day turnaround on technical video.

Client base of venture-backed devtool startups (Y Combinator, Initialized, Innovation Norway), with a published AppSignal case study claiming a 400% conversion-rate lift in three months.

A Y Combinator-backed docs studio for devtools with Supermemory, Greptile, Reducto and Rootly shipped, agents doing the code verification and humans doing the architecture.

Diátaxis expanded to fourteen content types, 27 rules, and six pluggable style guides — Google, Microsoft, Stripe, Canonical, Good Docs, or minimal.

Thirty-three checks across AI discoverability, structure, content quality, SEO and internal linking, returned as a 0–100 score with prioritised fixes.

Generates or audits a README against a cited standard, scored out of 100 across identity, trust, onboarding, structure, craft, and agent-readability.

Scores SDK docs across the six sections integration actually fails in — installation, quick start, error handling, troubleshooting, examples, best practices — as an HTML report.

Scaffolds and validates an Agent Skill — the packaging format that puts your conventions, gotchas and idiomatic examples inside the agent your users already work in.

Probes robots.txt, llms.txt and llms-full.txt, scores AI-readiness against a structured checklist, and needs no API key.

Reads git history, groups changes into added, improved, fixed and breaking, drops internal noise, and returns release notes a human can edit in ten minutes.

Anthropic's Connectors Directory is a gated, vetted distribution channel with no published review clock — a reminder that shipping an MCP server isn't the same as getting it in front of users.

The 2026-07-28 spec drops MCP's session handshake, hardens auth, and ships day-one across four SDKs with six hyperscalers behind it — the protocol crossed from experiment to backed standard, which resets both the positioning window and the migration bill.

In two weeks at least five teams shipped the same primitive — a policy gate between an AI agent and an irreversible action — and independent research is now naming it. The category is forming; the positioning window is open.

Sinch launched Agent Tools on 2026-08-04: VS Code/JetBrains/Open VSX extensions, an MCP server exposing live Sinch API definitions, a Simulator Mode that tests integrations before an account exists, and "Sinch Skills" — structured implementation guidance for AI coding assistants including Claude Code, Cursor, GitHub Copilot and ChatGPT Desktop. It joins a growing roster of vendors making their platforms "agent-safe by design" rather than just documented for humans.

Developer Josh Symonds documented a dated timeline of his MCP connector submission to Anthropic's directory: filed 2026-03-22, told as late as 06-15 it was "in the review queue," then informed on 07-30 the queue had been discarded and he would need to resubmit — against OpenAI approving the same connector for its own directory in 29 days. It is a single account, not an Anthropic-confirmed pattern, but it is the first concrete stranded-integrator story since the MCP spec went stateless.

Unblocked published a piece arguing that raw MCP connectors do not replace a curated context layer — that giving an agent access to every system is not the same as giving it the right context. It landed at 8 points with no comments, in the same week MCP support consolidated into a baseline expectation across every major vendor.

The 2026-07-28 Model Context Protocol spec drops `initialize` and session IDs so any server instance can answer any request behind a plain load balancer, graduates Tasks and MCP Apps into a versioned extensions framework, and hardens auth with RFC 9207 issuer validation in place of Dynamic Client Registration. All four Tier-1 SDKs shipped support on publication day, with AWS, Cloudflare, Figma, Google Cloud, Microsoft and Netlify publicly committed.

MCP's biggest spec revision started a twelve-month migration clock under every MCP server the same week GitHub retired Models on a six-week runway with no like-for-like replacement. How you end things is becoming as much of a marketing surface as how you launch them.

GitHub priced AI code review as a seat plus a meter on Monday, then shipped the dashboard that counts unused Copilot seats on Wednesday — the vendor selling AI seats is now arming buyers to find the shelfware, because the ROI-skeptical renewal conversation demands it.

GoDaddy and Atlassian both shipped product for the agent-as-user in the same week — the thread graduated from tiny Show HNs to platform roadmaps — while Cerebras showed how quietly sunsetting a free tier burns trust it took years to earn.

Machine-mediated discovery moved a step this week — from AI assistants reading your docs to agents comparing and picking vendors — while a zero-budget open-source launch topped Hacker News on reputation alone.

Machine-mediated discovery is real — Vercel gets 10% of signups from ChatGPT — but most of what's sold as GEO is ritual. The work that pays is docs and API surfaces you should want anyway.

Coding assistants (Claude, Cursor, Copilot) and answer engines increasingly read and cite developer documentation, and practitioners are restructuring docs — and adding llms.txt — so machines can retrieve them, not just humans.