# ScalablyAI > ScalablyAI (Scalably) builds and runs governed AI workforces inside real business operations. Its agents work through Slack, Telegram, WhatsApp, and the systems a company already uses. Scalably also publishes first-party implementation and engineering research across OpenAI Codex, Claude, agent SDKs, MCP, evaluations, and production operations. Machine-readable representations of selected articles are published under /_ai/ and served to AI agents at the canonical /blog/ URLs via content negotiation (Accept: text/markdown). ## Company - [What Scalably builds](https://scalably.io/): production AI agents for business operations - [AI agent blueprint](https://scalably.io/ai-workforce/map): free, ungated tool that designs the AI agent for one recurring workflow as a blueprint (trigger, sources, work, approval, destination, completion check) with a pilot plan; exports PDF, Markdown and JSON - [AI workforce](https://scalably.io/ai-workforce): the Company OS architecture, roles, controls, rollout, and metrics - [AI agent development company](https://scalably.io/ai-agent-development-company): the buyer standard for workflow contracts, evaluation, permissions, deployment, maintenance, and recovery - [Slack AI agents](https://scalably.io/slack-ai-agents): channel-native agents that execute work across company systems - [Platform](https://scalably.io/platform): how the agent platform works - [About](https://scalably.io/about): the company and founder - [Author: Pavle Lazic](https://scalably.io/author/pavle-lazic): founder, writes all technical articles from production experience - [Where to find Scalably](https://scalably.io/social): every official channel, package registry and feed, with a vCard and QR; agent version with the machine endpoints at https://scalably.io/_ai/social.md ## Business implementation - [AI agents for business](https://scalably.io/blog/ai-agents-for-business): how to select, scope, test, and operate the first workflow - [AI agents in agriculture](https://scalably.io/blog/ai-agents-in-agriculture): a privacy-safe field guide from work order to verified update and management decision ## Agent engineering - [Qwen Flash serving field report](https://scalably.io/blog/qwen-flash-next-dual-rtx-5090-serving-benchmark): completed256K serving cells, returning history, overload fairness and recipes. - [Context compaction made our agent blame the user](https://scalably.io/blog/context-compaction-agent-blamed-user): an SDK interrupt recorded as the user refusing a tool, why the transcript cannot say who interrupted, and the fix. - [Codex skills](https://scalably.io/blog/codex-skills): reusable workflows, trigger design, and deterministic checks - [Codex vs Claude Code](https://scalably.io/blog/codex-vs-claude-code): a first-hand workflow migration with the production runtime kept separate - [Claude Code hooks: a practical guide](https://scalably.io/_ai/blog/claude-code-hooks-guide.md): lifecycle events, exit-code semantics, and the production PreToolUse guard pattern - [The Claude Agent SDK: build your own agent](https://scalably.io/_ai/blog/claude-agent-sdk-guide.md): the agent loop, tools, MCP servers, SDK-vs-Claude-Code decision - [How to build an MCP server in TypeScript](https://scalably.io/_ai/blog/mcp-server-typescript.md): full walkthrough with the official TypeScript SDK - [MCP vs function calling](https://scalably.io/_ai/blog/mcp-vs-function-calling.md): what the Model Context Protocol adds over per-app tool wiring - [Claude Code best practices](https://scalably.io/_ai/blog/claude-code-best-practices.md): setup and usage patterns from running Claude Code across client accounts ## Skills Open-source Agent Skills derived from Scalably's production agents (https://github.com/scalably-io/agent-skills). For each skill, the raw SKILL.md is at https://scalably.io/skills/.md, the machine-readable page at https://scalably.io/_ai/skills/.md; the JSON index is at https://scalably.io/skills/index.json. - [research](https://scalably.io/_ai/skills/research.md): Use when the user requests research, a deep dive, a compare X vs Y, trend analysis… - [report](https://scalably.io/_ai/skills/report.md): Use when the user requests a report, analysis, summary, dashboard, breakdown, deep dive… - [automation](https://scalably.io/_ai/skills/automation.md): Use when the user requests bulk or repetitive data work that doesn't need judgment. - [prompt](https://scalably.io/_ai/skills/prompt.md): Rewrite and optimize prompts for the current Claude model. - [memory](https://scalably.io/_ai/skills/memory.md): Use when the user says remember, forget, recall, or asks what do you know about something, or wants to save a correction, preference… - [memory-system](https://scalably.io/_ai/skills/memory-system.md): The memory-system maintenance contract. - [runtime-constraints](https://scalably.io/_ai/skills/runtime-constraints.md): Use when uncertain whether a tool, package, or library is available in the current agent runtime, before any pip/npm/apt install… - [browser-scrape](https://scalably.io/_ai/skills/browser-scrape.md): Retrieve and extract public web-page content with Scrapling, escalating from a plain HTTP fetch to a real browser to an anti-bot stealth… - [publish-html](https://scalably.io/_ai/skills/publish-html.md): Use when the user wants a public URL for an HTML deliverable (report, dashboard, presentation) so they can share it by link. - [daily-log](https://scalably.io/_ai/skills/daily-log.md): Extracts a structured evidence ledger from today's session transcripts (actions, corrections, positive signals, workflow patterns… - [dream](https://scalably.io/_ai/skills/dream.md): Runs the nightly memory-consolidation routine defined by the memory-system skill. - [weekly-memory-cleanup](https://scalably.io/_ai/skills/weekly-memory-cleanup.md): Runs the memory-system skill's Sunday deep pass. - [friday-feedback](https://scalably.io/_ai/skills/friday-feedback.md): Runs a natural weekly check-in. - [workspace-reorg](https://scalably.io/_ai/skills/workspace-reorg.md): One-time deep cleanup of a project root that has accumulated loose files. - [anchor-policy](https://scalably.io/_ai/skills/anchor-policy.md): Validate anchor text placement, URL correctness, and SEO positioning in a guest-post draft against a target anchor spec. - [webmaster-policy](https://scalably.io/_ai/skills/webmaster-policy.md): Validate a guest-post draft against a target site's webmaster requirements. - [classify](https://scalably.io/_ai/skills/classify.md): Batch-classify a list of website domains by niche (primary + secondary + confidence). - [link-insertion-finder](https://scalably.io/_ai/skills/link-insertion-finder.md): Find the best article on a target site for inserting a backlink, given anchor text and a destination URL. - [batch-contact-email](https://scalably.io/_ai/skills/batch-contact-email.md): Extract a contact email for each domain in a list (no niche or quality filtering, assumes the domains are already vetted). - [lead-pipeline](https://scalably.io/_ai/skills/lead-pipeline.md): Runs a B2B lead-generation pipeline end to end. - [internal-linking](https://scalably.io/_ai/skills/internal-linking.md): Internal-linking campaign engine. - [guest-post-writer](https://scalably.io/_ai/skills/guest-post-writer.md): Write a complete guest-post article from a brief (target site, topic, target keyword, anchor and URL). ## MCP servers MCP servers we run in production: 8 built by Scalably and published under io.scalably, plus the third-party servers we run with the versions we pin. For each server the machine-readable page is at https://scalably.io/_ai/mcp/.md, the JSON record at https://scalably.io/mcp/.json; the index is at https://scalably.io/mcp/index.json. - [Google Search Console MCP](https://scalably.io/_ai/mcp/gsc-mcp.md): Read-only Google Search Console MCP: Search Analytics, sitemaps, URL inspection. - [Google Analytics 4 MCP](https://scalably.io/_ai/mcp/ga4-mcp.md): Google Analytics 4 MCP: reports, realtime, funnels, pivots, metadata, admin lists. - [Google Ads MCP](https://scalably.io/_ai/mcp/google-ads-mcp.md): Google Ads MCP: GAQL queries, resources, recommendations, keyword ideas, change history. - [Hunter MCP](https://scalably.io/_ai/mcp/hunter-mcp.md): Hunter.io MCP: domain search, email finder and verifier, enrichment, discovery, leads. - [Klaviyo MCP](https://scalably.io/_ai/mcp/klaviyo-mcp.md): Klaviyo MCP: campaigns, flows, segments, metrics, reports, profiles across many accounts. - [Figma MCP](https://scalably.io/_ai/mcp/figma-mcp.md): Figma MCP: list frames in a file and fetch one frame's JSON and image safely. - [DataForSEO MCP](https://scalably.io/_ai/mcp/dataforseo-mcp.md): DataForSEO MCP wrapper: searchable API docs plus normalized calls to the official server. - [Shopify MCP](https://scalably.io/_ai/mcp/shopify-mcp.md): Read-only Shopify Admin GraphQL MCP: query, introspect, bulk export, ShopifyQL. - [Tavily MCP](https://scalably.io/_ai/mcp/tavily-mcp.md): Web search, extract, map and crawl for research agents. - [DataForSEO MCP Server](https://scalably.io/_ai/mcp/dataforseo-mcp-server.md): The upstream server our dataforseo-mcp wrapper runs; install it directly if you do not need the docs index. - [Microsoft 365 MCP Server](https://scalably.io/_ai/mcp/ms-365-mcp-server.md): OneDrive and Microsoft Graph file access behind a per-tenant folder policy. - [Penpot MCP Server](https://scalably.io/_ai/mcp/penpot-mcp-server.md): Self-hosted Penpot design files for agents that produce layouts. - [lemlist MCP (remote)](https://scalably.io/_ai/mcp/lemlist-mcp.md): Outreach campaign data from lemlist's hosted MCP, bridged over stdio with mcp-remote. - [MCP Toolbox for Databases (BigQuery)](https://scalably.io/_ai/mcp/mcp-toolbox-bigquery.md): BigQuery datasets, tables and SQL for reporting agents, through a tools.yaml config. - [Scrapling MCP](https://scalably.io/_ai/mcp/scrapling-mcp.md): Stealth fetching and Cloudflare-aware scraping behind our browser-scrape skill; we run it under a fail-closed guard. - [Playwright MCP](https://scalably.io/_ai/mcp/playwright-mcp.md): Browser automation and QA from Claude Code and Codex on the operator's machine. - [Linear MCP (remote)](https://scalably.io/_ai/mcp/linear-mcp.md): Issue and project tracking for every engineering session. - [Context7](https://scalably.io/_ai/mcp/context7-mcp.md): Current library documentation before any version-sensitive claim or dependency change. - [Codex CLI as an MCP server (retired)](https://scalably.io/_ai/mcp/codex-cli-mcp.md): Was a second implementer and reviewer reachable as a tool from Claude Code, until OpenAI removed codex mcp-server. ## Agent hooks Guardrails that run in front of a coding agent, open source at https://github.com/scalably-io/agent-hooks. For each hook, the raw script is at https://scalably.io/hooks/.sh and the machine-readable page at https://scalably.io/_ai/hooks/.md; the JSON index is at https://scalably.io/hooks/index.json. - [bash-guard](https://scalably.io/_ai/hooks/bash-guard.md): PreToolUse (Bash) hook: triage destructive shell commands before they run. - [verify-before-commit](https://scalably.io/_ai/hooks/verify-before-commit.md): PreToolUse (Bash) hook: run a project's own verification before commit or push. - [guard-agent-nesting](https://scalably.io/_ai/hooks/guard-agent-nesting.md): PreToolUse hook: one level of delegation. - [guard-agent-background](https://scalably.io/_ai/hooks/guard-agent-background.md): PreToolUse hook: a spawned subagent must run in the foreground, explicitly. - [web-fetch-guard](https://scalably.io/_ai/hooks/web-fetch-guard.md): PreToolUse and PostToolUse hook (WebFetch): stop asking a site that has already refused. - [stuck-detector](https://scalably.io/_ai/hooks/stuck-detector.md): PostToolUseFailure hook: catch an agent retrying the same failing call. - [pre-compact-state](https://scalably.io/_ai/hooks/pre-compact-state.md): PreCompact hook: write a small state snapshot before the context is compacted. - [tidy-scan](https://scalably.io/_ai/hooks/tidy-scan.md): Stop hook: surface untidy repository state at the end of a turn. ## All articles - [Blog index](https://scalably.io/blog/): all published articles (HTML) ## Research - [Fast Memory, Slow Weights: A Production Architecture for Self-Improving AI Agents at the System Layer](https://scalably.io/research/fast-memory-slow-weights): Production field report (v1.0, 2026-09-20). A rejected LoRA, a fast scoped reversible memory loop, a slow gated weight loop, evidence rows and 29 verified citations. PDF: https://scalably.io/research/Scalably_Fast-Memory-Slow-Weights_v1.0.pdf - [Move the energy, not the data centre? A falsifiable model of orbital energy for AI](https://scalably.io/research/orbital-power-fabric): Falsifiable model (v1.0, 2026-09-27). Microwave beaming, orbital mirrors and GPUs in orbit tested against ground solar on launched mass and launch cost, a photonic inference negative result, an open photonic test-chip layout and a public quote-per-claim evidence ledger. Built with AI agents, reviewed by AI models. PDF: https://scalably.io/research/Scalably_Orbital-Power-Fabric_v1.0.pdf Code and data: https://github.com/scalably-io/orbital-power-fabric - [Is the graph doing anything? A sourced survey of recurrent self-routing neural graphs as an alternative to layered networks](https://scalably.io/research/neural-graph-substrate): Sourced survey (v0.5, 2026-10-01). Whether a persistent self-routing recurrent graph gives more reasoning per stored parameter than a weight-shared looped Transformer; what is already published, what is open, and a preregistered matched protocol that has not been run. Built with AI agents, reviewed by AI models. PDF: https://scalably.io/research/Scalably_Neural-Graph-Substrate_v0.5.pdf Code and data: https://github.com/scalably-io/neural-graph-substrate