Tasklet Review

$25/month gets you 10,000 credits, unlimited teammates, no seat tax. That pricing shape matters more than the agent demo — most competitors in this lane still charge per user, which punishes exactly the team collaboration Tasklet is selling.

The browser-based fallback for tools without an API is a real differentiator, not a gimmick. Vendor portals and ancient internal tools are where automation projects usually die, and letting an agent click through them like a human is a genuine capability, not a checkbox.

SOC 2 attestation and per-agent credit caps show someone thought about admin fear before launch. My hesitation: usage-based billing on autonomous 24x7 agents can run up fast if guardrails aren't tuned, and I couldn't find granular per-unit rates beyond the three named tiers. Pilot one workflow, watch the credit burn for 60 days, then decide.

The isolated-cloud-per-agent design is the tell here. Most agent platforms run everything in a shared execution pool and bolt on tenant isolation later as a compliance patch; Tasklet built it in from the start, and SOC 2 attestation backs that up. That's the right call if you're routing Stripe, QuickBooks, and Zendesk credentials through shared connections — the blast radius of a single agent misbehaving stays contained.

The browser-fallback for API-less tools is the more interesting architectural bet. It means Tasklet isn't just an integration layer over documented APIs — it's making a claim on legacy internal apps and vendor portals nobody's ever going to build a clean API for. If that holds up under real vendor-portal UI drift, it's a durable moat. If it's brittle, you inherit a maintenance tax on every workflow that touches it.

Three years out: instant apps and shared context turn Tasklet into a place where institutional workflow logic lives, not just a task runner. That's a real dependency to weigh — migrating off a platform that's absorbed your team's tribal knowledge is a different problem than swapping a chatbot vendor.

Three named tiers: Starter $25/month for 10,000 credits, Pro $100/month for 40,000 credits, Custom $250/month for 100,000 credits. Custom amounts run $25 to $10,000 in $5 increments. No per-seat fee — unlimited team members share one pool. Fine for procurement, bad for forecasting: usage scales with agent activity, not headcount, so a busy quarter can blow past last quarter's bill with no seat count to warn you.

Credits-to-dollars conversion isn't published beyond the three tiers. That's the gap. $100/month buys 40,000 credits — what one drafted email or one lead enrichment costs in credits, I couldn't find. Year 3 TCO depends entirely on agent volume, and volume depends on adoption you can't predict at signing.

Enterprise tier is demo-gated, standard for SOC 2-compliant platforms selling into security-conscious buyers. No auto-renewal terms or contract length listed on the pricing page — worth asking before signing anything beyond Starter.

The pitch that lands technically: agents connect to thousands of SaaS tools or any MCP server, and when there's no API, they drive a built-in browser against the vendor portal like a human would. That's the unglamorous integration work every automation engineer eventually hits — the internal app with no docs, the vendor that never shipped an API. Handling it natively instead of forcing a workaround through some other tool is a real design choice.

Day-3 concerns are less about capability and more about observability. Isolated per-agent cloud environments plus SOC 2 compliance cover the security box, and per-agent credit caps give you a spend ceiling to watch. But I couldn't find API reference docs in my pass over the site — for a platform built around 'any API, including private APIs,' that's the exact page a practitioner needs before trusting it with a Stripe or QuickBooks workflow.

Multi-model routing (Claude, GPT, Gemini, Kimi) is genuinely useful for tuning cost per task. Tradeoff: browser-driven steps are inherently slower and more brittle than a clean API call — that's the cost of the fallback, not a flaw unique to this platform.

You describe a goal, connect Gmail or HubSpot or Stripe, and an agent goes and does the thing. That's the pitch, and the feature list backs it up: instant apps, shared context, per-agent credit caps starting at $25/month for 10,000 credits. What's missing from what I could find is the texture of using it daily. What does the agent say when it's stuck halfway through a task waiting on your approval? Does it sit there politely or does it look broken?

The Mac app is a good sign someone's thinking about how this lives on your desktop day to day, not just as a browser tab you forget about. But I couldn't find much on what happens when a browser-driven step against an old vendor portal times out mid-run, and that's the moment that actually tests a tool like this.

The admin console and SOC 2 badge suggest the backend is taken seriously. Whether the front-of-house moments feel that solid too, I genuinely don't know yet.

"Run your business on agents" is a big promise for a product whose cheapest tier is $25/month for 10,000 credits. What's a credit worth in actual tasks? Not stated anywhere I could find. That gap between headline ambition and pricing-page vagueness is the first thing I'd want closed before rollout.

The capability list — instant apps, browser fallback for API-less tools, SOC 2, multi-model routing — reads specific, not aspirational. That's a point in its favor. I've seen agent platforms promise 'connects to anything' with three integrations live. Tasklet names Gmail, HubSpot, Zendesk, Stripe, QuickBooks specifically, which is the kind of detail that either holds up or gets quietly walked back.

Custom tier scales to $10,000/month with no stated ceiling above that, which tells me enterprise deals are handled off-page, standard for this category. Fine if you're negotiating. Less fine if you just want a number to plan around.