Sacra estimates that Tasklet hit $10M in annual recurring revenue (ARR) in May 2026, up from ~$385K at the end of 2025.

In January 2026, Tasklet was adding more monthly recurring revenue than its predecessor product Shortwave had accumulated across its entire first four years of operation.

Growth appears to be largely product-led, with a free tier that includes functional utility and a low-friction upgrade path as users expand their automation workloads. The customer base today skews toward early-adopter individuals, power users, and small teams, a cohort that has historically seeded larger workflow automation businesses before enterprise sales efforts begin.

Tasklet raised a $20M round in April 2026 at a $175M post-money valuation, led by Union Square Ventures and Lightspeed.

The round included participation from Y Combinator. Tasklet joined YC's Spring 2026 batch, founder Andrew Lee's second time in the program, fifteen years after Firebase went through YC S11, as well as angels including Jeff Dean and Patrick and John Collison.

Tasklet appears to have been bootstrapped or minimally funded through the Shortwave years before pivoting to Tasklet in late 2024. Total disclosed primary equity raised stands at $20M.

Tasklet is a cloud-based AI agent platform where users describe a task in plain English and the agent determines how to complete it, connects to the relevant tools, and runs it automatically on a schedule, without a workflow builder, node graph, or trigger-action rules.

The interface centers on a chat window. A user might type something like: pull together a daily briefing from my calendar, inbox, and key Slack channels, summarize anything urgent, and send it to me at 7am. The agent reads that instruction, selects the services to connect, builds a plan, and starts running it. The user reviews the output, gives feedback in plain language, and the agent adjusts. There is no separate setup mode, because the same conversational interface is used while the agent executes scheduled runs in the background.

Under that interface, Tasklet has four capability layers. The connections layer supports integrations with services like Gmail, Slack, Notion, HubSpot, Salesforce, Linear, and BigQuery through pre-built connectors. For services without a native connector, users can point Tasklet at any HTTP API with custom auth headers and the agent determines how to use it. For services with no API, Tasklet can spin up a cloud browser and interact with the website directly, extending access to data sources and tools across the internet.

Automations can be activated through scheduled triggers (daily, weekly, custom cron), email triggers, webhook triggers from external services, and Slack message triggers, so the system can run without a human initiating each task. When a workflow requires bulk data processing or number-crunching, Tasklet writes and executes code in an isolated cloud sandbox, handling transformations that would otherwise require a technical workflow engineer.

Instant Apps, added in early 2026, lets the agent generate a deployed web UI on demand, such as a dashboard, form, time tracker, or data entry interface, connected to live business data. A user can describe the interface they need in a single prompt and get a working app in minutes, with the agent inferring the relevant data sources and adding interaction elements without explicit instructions.

Tasklet sells to individuals and small teams on a credit-based subscription model with four published tiers: a free plan with daily bonus credits, a $25/month Starter plan, a $100/month Pro plan, and a Custom plan starting at $250/month for higher-volume users.

Credits are consumed based on task complexity, context size, the number of tools active in a run, trigger frequency, and the intelligence level selected. Users choose between four intelligence tiers, Basic (Claude Haiku), Advanced (Claude Sonnet with adaptive thinking), Expert (Claude Opus), and Genius (Claude Opus with enhanced context), and more capable agents consume credits faster. That creates an embedded upsell path: as users assign more complex or higher-stakes work to agents, they move toward higher intelligence levels and larger credit pools.

The primary cost driver is Anthropic API consumption. A single complex multi-step agentic run using Claude Opus with extended context and multiple tool calls carries meaningful inference costs, and gross margins depend heavily on caching efficiency and the distribution of intelligence levels across the user base. Prompt caching, a technique the team developed during the Shortwave years, is a key lever for keeping per-run costs manageable as session context grows longer.

Go-to-market is entirely product-led, with a free tier that has real functionality, a low-cost entry point, and an escalation path as workloads deepen. Enterprise pricing is available on request, but enterprise-grade features like SSO, audit trails, and org-level cost controls are roadmap items rather than generally available today, which limits the current upmarket motion.

The automation and AI agent market in 2026 spans three competitive clusters, each addressing the same underlying problem, connecting AI reasoning to business tools, from a different direction. Tasklet competes most directly with AI-native agent builders in the near term, while longer-term pressure comes from workflow incumbents adding AI and foundation model providers moving into the action layer.

Zapier, Make (formerly Integromat), and n8n define the incumbent layer. Zapier has over 8,000 integrations built over fifteen years, along with enterprise adoption among buyers that prioritize predictable, auditable, deterministic automation.

All three have responded to the agentic shift by adding AI on top of existing workflow infrastructure, Zapier Agents in beta, Make's Maia, and n8n's 70-plus AI nodes with native LangChain integration. The structural difference from Tasklet is that in these tools, the human still defines the workflow graph and the AI executes discrete steps within it.

Tasklet's core claim is that this hybrid approach is transitional rather than the end state, and that recent gains in model capability reduce the need for the workflow graph itself. For buyers already invested in Zapier with complex existing automations, switching costs are real. For new users starting fresh, the question is whether they adopt a node graph at all.

Lindy is the closest analog on product philosophy: a personal AI assistant with pre-built personas for email, calendar, and meeting management, SOC2 and HIPAA compliance, and pricing that starts at roughly $50/month. Relevance AI is aimed more at technical teams building custom multi-agent systems, functioning more as a developer platform than an end-user product. Sauna, Fyxer, and Zo Computer focus on chiefs of staff, managers, and executive assistants, a narrower vertical slice of the same knowledge-work automation thesis.

Relative to this cluster, Tasklet's product scope is broader: cloud-native 24/7 execution, computer-use fallback for services without APIs, a code execution sandbox, and Instant Apps. That gives it a wider surface area than most persona-based competitors offer today, while also making the product more general-purpose.

The largest strategic pressure comes less from Zapier or Lindy and more from Anthropic, OpenAI, and Google building agentic products on top of their models. Claude already has tool-use capabilities. OpenAI launched Operators. Google has Gemini with function calling. OpenAI's acquisition of OpenClaw in March 2026 indicates direct intent to move down-stack into the action layer.

Tasklet's near-term moat is the connections layer: production-quality integrations for thousands of services, OAuth flows, reliability infrastructure, and computer-use for API-less services. Building that layer is difficult, and the Firebase team's experience with scalable cloud infrastructure is relevant. That said, it is unlikely to remain defensible indefinitely if a model provider invests aggressively in the same layer using its existing distribution.

Tasklet's current $5M ARR base sits almost entirely in the prosumer and SMB segment. Its expansion logic runs in three directions: moving upmarket into enterprise, expanding the product surface into a broader operating layer for knowledge work, and capturing integration infrastructure spend that today flows through iPaaS players.

The knowledge-work automation market at the enterprise tier includes Zapier, Workato, Tray.io, and the native agent tooling bundled into Salesforce Agentforce, Microsoft Copilot Studio, and AWS Bedrock AgentCore. These buyers typically spend more per seat and show longer retention once a workflow platform is embedded in operations.

Tasklet's path into this segment depends on shipping enterprise features it currently lists on its roadmap, including SSO, audit trails, org-level cost controls, and compliance certifications. The Firebase team has built enterprise-grade infrastructure before, and the $20M raise gives it capital to close this gap. If those features ship, the $25–$100/month tiers can serve as a bottom-up wedge into organizations where power users already run Tasklet automations.

The Instant Apps feature points to a broader expansion: replacing work that previously required a developer with a chat prompt. Today it generates dashboards, forms, and data-entry interfaces connected to live business data. The longer-term opportunity is situational software, custom UIs built for a specific workflow, used for weeks or months, then discarded or replaced, generated on demand instead of commissioned from an engineering team.

That would move Tasklet beyond automation into lightweight application creation, competing at the edge with no-code tools like Retool and other internal-tool builders. The internal tooling and business application development market is large, and generating a working app from a text description rather than a drag-and-drop canvas is a distinct product approach.

As AI agents spread across the enterprise, the integration fabric connecting them to business systems becomes more valuable. Today that market includes Zapier, Workato, Ampersand, Alloy, Merge, and Finch, companies that own auth, data mapping, sync, and the developer-facing integration layer.

Tasklet says it can interface with arbitrary APIs and services without pre-built connectors, including through computer use for services with no API at all. That challenges the premise that integration infrastructure requires years of connector-building. If that claim holds at enterprise reliability standards, Tasklet would compete not just with workflow automation vendors but with the broader iPaaS and integration middleware market, a segment that has historically supported large, durable businesses.

Model dependency: Tasklet runs almost entirely on Anthropic's Claude across four model tiers, so any change in Anthropic's pricing, availability, or terms of service flows directly through to Tasklet's cost structure and product reliability, and that dependence becomes more acute as the company moves upmarket and customers demand model optionality.

Foundation model encroachment: The connections layer and trigger infrastructure that constitute Tasklet's current moat could be replicated by Anthropic, OpenAI, or Google if any of those providers decides to build the same integration depth with the distribution advantages of an incumbent model platform, compressing Tasklet's differentiation before it has accumulated sufficient enterprise switching costs.

Enterprise readiness gap: Tasklet's absence of generally available SSO, audit trails, org-level cost controls, and compliance certifications means the company cannot yet pursue the enterprise segment where automation platforms generate their most durable and highest-value revenue, and a nine-person team scaling rapidly to close that gap while maintaining product quality faces execution risk.