Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
Retool is the leading low-code platform for building internal tools, now expanding aggressively into AI-native capabilities. The core platform lets developers build admin panels, dashboards, and operational workflows using drag-and-drop components connected to any database or API. Over 10,000 companies use Retool, including Amazon, DoorDash, NBC, and Stripe.
The AI expansion (2025) adds two major capabilities: AppGen generates full internal apps from natural language prompts, and Agents creates autonomous AI workflows powered by configurable LLMs with MCP server integration. This positions Retool as both a traditional internal-tool builder and an emerging AI agent orchestration platform.
Differentiator: Unlike pure-play AI coding tools, Retool provides governed infrastructure for deploying AI agents that interact with enterprise data sources, with self-hosted deployment, audit trails, and RBAC built in from the platform's enterprise heritage.
Best fit: Enterprise teams building internal tools and operational workflows who want to augment existing Retool deployments with AI agent capabilities. Less suitable for teams seeking general-purpose code generation or IDE-integrated coding assistance.
Adoption & Proof Points
- Funding: Total raised ~$350M. $45M Series C (July 2022) at $3.2B valuation, led by Sequoia Capital — now 4 years stale, no new round announced. Investors include Stripe founders (Patrick and John Collison), Nat Friedman, Elad Gil, BOND Capital.
- Revenue: $200M+ ARR — crossed in 2025 (up from $120M Oct 2025 / $90M end-2024). Cashflow breakeven while growing rapidly (CEO David Hsu). Agents priced hourly (wall-clock; ~$50/mo free allowance, ~$40/hr for GPT-4.1-class models) — a new cost-unpredictability vector.
- Scale: 25,000+ companies; 100,000+ internal apps built monthly. Enterprise customers include Amazon, DoorDash, NBC, Stripe; regulated-industry references Plaid and Hims (self-hosted VPC). Retool's Feb 2026 Build-vs-Buy report claims 35% of enterprises have replaced SaaS with custom software.
- Headcount: ~416 (Apr 30 2026), down from 460 (Jan) / 451 (Feb) following a ~9% layoff in April 2026 (entire customer-success-manager team plus recruiting and workplace teams) — framed as efficiency/cost reallocation, not financial distress (company is cashflow breakeven with growing ARR). Leadership maturing: founder-CEO David Hsu, plus first COO (Mark Schaaf), CFO (Adrian Kaplan), CRO, CPO, and Head of Security.
- G2 reviews: Consistently rated 4.5+/5. Users praise speed of internal tool development and breadth of integrations (one dev does the work of a team). Common complaints: per-user pricing expensive at scale (a 50-person team with 200 viewer seats ≈ $66K/yr), SSO/self-hosted gated behind higher tiers, editor lag on large apps and slowness with many concurrent users, steep learning curve for advanced features.
- Market position: Category leader in low-code internal tools, but facing existential AI-native-builder disruption — Lovable ($300M ARR <18 months), Bolt, v0 generate ownable, deployable code (with security tradeoffs: ~45% of AI-generated code has vulnerabilities; Lovable RLS exposure incident). Retool's counter is governance-as-moat ("Code is free. Now what?", Mar 2026). Primary low-code competitors: Appsmith (open-source, self-host), Superblocks (server-side execution for large apps), ToolJet, Budibase, DronaHQ, UI Bakery.
Recommended Use Cases
- Enterprise teams building governed internal tools (admin panels, dashboards, approval workflows)
- Organizations augmenting existing Retool deployments with AI agent capabilities
- Teams needing AI-powered operational workflows connected to enterprise data sources
- Companies requiring self-hosted deployment for data sovereignty or compliance
- Non-developer business users generating internal apps from prompts via AppGen
Risks & Limitations
- AI Agents/Assist in beta: Public beta since July 2025, no GA, no production SLA, no independent reliability benchmarks. Enterprise adoption of agent features is early and unvalidated at production scale.
- AI performance/reliability complaints (expanded 2026): Multiple community-forum threads report AI Assist extreme slowness (30+ min on simple tasks), browser crashes, frozen prompt windows, and apparent code corruption; a separate thread reports very slow AI Agents. Concentrated on the beta AI features.
- Layoff (Apr 2026): ~9% workforce cut (entire CS-manager + recruiting + workplace teams; headcount 460→416). Framed as efficiency at a cashflow-breakeven, $200M+-ARR company — a viability headwind, not financial distress.
- AI-native disruption: Lovable/Bolt/v0 generate ownable, deployable code and are growing explosively (Lovable $300M ARR). Retool's moat is governance + enterprise infrastructure, not code generation.
- Pricing complaints: Per-user pricing expensive at scale (Business commonly cited ~$50–65/user/mo; 50-person team w/ viewers ≈ $66K/yr). SSO requires Business+, self-hosted requires Enterprise. Agent hours (wall-clock billing) introduce new cost unpredictability.
- Self-hosted now Enterprise-only (Feb 2026): Reduces governance accessibility for mid-market; capability intact for large/regulated enterprises.
- 2023 security breach: 27 cloud customers compromised via SMS phishing; resolved with post-mortem; self-hosted unaffected. Demonstrates social-engineering risk surface. (May 2026 NGINX CVE-2026-8711 is a transitive self-hosted dependency-patch request, not an unpatched critical Retool vuln.)
- Platform lock-in: No code export — apps are tied to the Retool runtime. Unlike Lovable or Bolt, code cannot be extracted and deployed independently.
- Not a coding tool: Retool builds internal tools, not general software. AI features augment the platform but do not generate portable, production-grade code.
- Self-hosted upgrade friction: Community reports of stuck DB migrations and failed version upgrades; editor lag on large apps and slowness under high concurrency.
Capabilities & Integration
Platform capabilities: Drag-and-drop UI builder with 100+ pre-built components. Native connectors to PostgreSQL, MySQL, MongoDB, REST APIs, GraphQL, gRPC, S3, Firebase, and 50+ data sources. JavaScript and SQL for custom logic. Workflows engine for backend automation with cron scheduling and webhooks.
AI features (2025): AppGen generates complete multi-page apps from text prompts. AI-powered query generation assists with SQL and JavaScript. Retool Agents (public beta July 2025) provides autonomous LLM-powered workflows with tool use, multi-step reasoning, and human-in-the-loop controls.
Agent architecture: Agents use an orchestration layer atop Workflows, with agent loops, predefined and custom tools, and first-class chat UX. Agents can access saved queries, workflows, MCP servers, and chain to other agents. Multi-model support — users choose their preferred LLM. Built-in evals for monitoring agent performance and costs.
Models: Multi-model — users choose the LLM. Claude Opus 4.7 (via Anthropic and Amazon Bedrock), GPT-5.4, GPT-4 Turbo, GPT-4.1 Nano, and Google Vertex AI available for AI resource queries, Agents, and Assist (as of Apr 2026).
MCP integration: Connect external MCP servers (GitHub, Slack, Stripe, Docker, Cloudflare) as agent tools. Protocol support up to version 2025-06-18. Retool's own MCP server (public beta) lets external agents manage the Retool org (write queries, manage users/invites, audit access, inspect resources/configs) — read/admin-focused today; app building and editing are on the roadmap, not GA.
Source Control: Native API-based Git integration with GitHub, GitLab, AWS CodeCommit, Bitbucket, and Azure Repos (expanded). Branch-based development with PR workflows — changes only via pull requests on branches, no direct writes to main/master; named point-in-time commits; code review via PRs. Multi-element branching for parallel development across apps, modules, and queries. React adoption (Mar 2026) lets external AI tools (Claude, Cursor, Codex) generate apps onto the governed platform.
Deployment: Cloud-hosted (default) or self-hosted (Docker, Kubernetes/Helm, ECS, Azure) — self-hosted now Enterprise-only (Feb 2026). Recent stable releases through May 13 2026 (3.334.x / 3.300.x); Workflows can select Python 3.10 or 3.14. Staging and release management. Custom domains.