AccessioAI @Accessio_AI

1/ Bulk operations on AI-credit-priced features used to be "fire and hope". New tool: get_quota. Read-only. Zero credits to call. ↓ 2/ Returns: plan → "pro" monthlyLimit → 2000 used → 1140 remaining → 860 resetsAt → "2026-07-01T00:00:00Z" 3/ You: "I want to re-scan 50 pages tonight. Do I have budget?" Claude calls get_quota, divides remaining by per-scan cost, tells you 43 will run now, 7 will queue past the reset. Stop guessing. Start preflighting. 4/ Scope: quota:read. Same data at /v1/quota for non-MCP callers.

keyboard_arrow_left Previous keyboard_arrow_right Next

1/ One Accessio.AI key, six possible scopes: scans:read scans:write alt-text:read alt-text:approve embed:read quota:read No wildcards. ↓ 2/ CI bot: scans:write + quota:read. Can re-run scans. Can read budget. Cannot list scans. Cannot read alt-text. Cannot mutate suggestions. Cannot read embed metadata. 3/ Reviewer agent: alt-text:read + alt-text:approve. Can review + approve. Cannot queue scans. Cannot read scans. Cannot see budget. Hand it out per-reviewer. Rotate when the reviewer rolls off. 4/ Scope denials write DENIED audit rows. Escalation attempts are visible, not silent. OAuth 2.1 with PKCE + dynamic client registration ships next quarter.

keyboard_arrow_left Previous keyboard_arrow_right Next

1/ Every agent action on Accessio.AI writes an audit row. What's new: each row is paired across services. 2/ A scan-run hits two services — the Node gateway and the Java scanner that does the AI work. Both write audit rows. Both pinned to the same requestId. SELECT action, service, status FROM audit_log WHERE request_id = '7c2f04a9-...'; → mcp.run_scan NODE SUCCESS → scans.run.requested JAVA SUCCESS 3/ One row in the dashboard. One filter in your SIEM. SOC2 stops being a forensics exercise. Args are SHA-256 hashed. IPs /24 truncated. No raw HTML. No PII.

keyboard_arrow_left Previous keyboard_arrow_right Next

1/ Reviewing 200 AI-generated alt-text suggestions used to mean clicking through 200 dashboard rows. Now: chat. ↓ 2/ "List pending alt-text suggestions for /products/runners-spring-red. Approve the ones that read naturally; reject the ones that just say 'image' or 'photo'." Claude calls list_alt_text → approve_alt_text → reject_alt_text in a loop. 3/ Pilot merchant last week: 200 suggestions in 4 minutes. Same site, same person, 38 minutes the month before via the dashboard. Rejection reasons feed the next generator batch. Loop tightens. 4/ Scope: alt-text:approve. Available on every Accessio.AI plan. mcp.accessio.ai/mcp

keyboard_arrow_left Previous keyboard_arrow_right Next

1/ For teams that want HTTP, not chat: api.accessio.ai/v1 is live. OpenAPI 3.1 spec at /v1/openapi.json. ↓ 2/ The surface: GET /v1/scans list, paginated GET /v1/scans/:id single session POST /v1/scans/:id/run queue re-run GET /v1/alt-text paginated POST /v1/alt-text/:id/approve POST /v1/alt-text/:id/reject GET /v1/embed-scripts GET /v1/quota 3/ Idempotency-Key on writes — a flaky CI retry doesn't double-spend AI credits. Same key within 24h = cached response replayed. 4/ Same scopes, same audit rows as our MCP. Pick the transport that fits your stack. Both live now.

keyboard_arrow_left Previous keyboard_arrow_right Next

1/ Shopify shipped Storefront MCP. Salesforce shipped a CLI MCP. Anthropic + OpenAI both shipped directories. In 2026, your SaaS isn't shippable unless an AI agent can reach it. We built Accessio.AI's surface in 6 engineer-weeks. ↓ 2/ Live now at mcp.accessio.ai/mcp: → 8 MCP tools across scans, alt-text, embed scripts, quota → OAuth 2.1 (PKCE, RFC 8707 audience binding) → Audit log paired across Node + Java services → Scoped API keys, no wildcards 3/ Tell Claude: "Show me my 10 most recent scans on accessio.ai. Re-run the one on /collections/spring-2026 — I just deployed a fix." Claude calls list_scans → run_scan. You watch the requestId update in the dashboard. Done. 4/ Submitted to: - Anthropic Connectors Directory - OpenAI Apps Directory Once approved, one-click install from inside claude.ai / chatgpt.com. Series on what we shipped + what we gated → /8 in the replies over the next 2 weeks.

keyboard_arrow_left Previous keyboard_arrow_right Next

Automated tools catch maybe 30% of real barriers. The rest needs human judgment and screen reader testing. Baking accessibility into every sprint beats a last-minute audit every time. #a11y #WCAG

WebAIM @webaim

2 months ago

#Accessibility testing works best when it’s ongoing and shared across teams, not saved for the end. This article from @TetraLogical explains why automated tools aren’t enough and why combining them with manual testing leads to better outcomes. tetralogical.com/blog/2026/01/0… #a11y

1 1 4 196 0

Most accessibility tools need a dev team and weeks of setup. AccessioAI takes 5 minutes. One script tag. Zero code changes. Works on Shopify, Wix, or any custom site. accessio.ai #accessibility #WCAG #a11y

@Equidox_A11y This is important context most businesses miss - meeting just the federal baseline can still leave you exposed in states like California or New York.

@accessnovaa AI is closing that gap fast, especially for alt text and reading order where context-aware models now outperform basic tagging tools.

@ranahson The fragmented enforcement is the real challenge here - businesses selling across borders need to meet the strictest interpretation, not just their local one.

On May 5, 1999, W3C, in cooperation with the WAI initiative, issued the first recommendation for a set of Web Content Accessibility Guidelines (WCAG 1.0). #InternetHistory

♿ Built an Accessibility Auditor as an AI agent on @arc Testnet! What it does: - WCAG 2.1 accessibility audit for any website - x402 payments in USDC - ERC-8004 on-chain agent identity (#963) - Telegram bot + REST API As a blind developer, I build tools that make the web accessible for everyone. #BuildOnArc #BuildOnCircle

Your Shopify store gets traffic — but can everyone navigate it? AccessioAI adds WCAG 2.2 compliance in one click. No code changes needed. accessio.ai #accessibility #Shopify #WCAG

@Equidox_A11y Prioritization is key — most teams stall because they try to remediate every PDF at once instead of focusing on the highest-impact documents first.

Weekly Reading List tpgi.com/weekly-reading… #accessibility #a11y #web #design #development #news

Share your slides and reports before the meeting—small step, big impact. My three visually impaired best practices improve presentations for everyone. wix.to/0kQpvhN #Accessibility #PresentationTips #A11y

WebMCP makes your site readable by AI agents - not just browsers. If an AI assistant can't navigate your site, you're invisible to a growing share of users. accessio.ai #accessibility #a11y #WCAG

@jecture_a11y Framing it as a consumer right changes everything - it moves accessibility from a nice-to-have into a legal baseline.

@A11ytestAI France setting the precedent means other EU member states will follow fast - the compliance window is shrinking.