What Tasklet AI is, and what an open‑source version changes

What Tasklet is, in its own words

Tasklet sells itself as "Automate any business task with AI agents that run 24x7." The claim is larger than a chat window: the agents own a job and keep it running while nobody watches. You write the task and connect the company tools. The agent then runs on a schedule, or when something happens: a new email, a Slack message, a row in the CRM.

The site claims every job function, and it puts named customer stories behind the claim. A Houston manufacturer runs Tasklet across four departments. A four-person marketing team took its Meta ads from three campaigns to sixty and brought the outsourced work in-house.

How the agents run and what they can reach

Tasklet is a managed cloud product: agents run in Tasklet’s isolated cloud environments and keep working after you close the app. Each agent gets its own environment and permission boundary. The credentials for its connections sit in an encrypted vault on Tasklet’s side. From there it reaches thousands of SaaS tools, MCP servers or any API, private ones included. It can also open a browser on a site with no API and run scripts in a command-line environment.

The approval layer is where the product earns its keep. In Tasklet’s words: "Decide what your agents can do on their own, and what they need permission for. They’ll reach out when they need your approval." An agent works unattended until it reaches something a person should see. Then it asks, and the answer comes back into the same cloud.

What each Tasklet tier costs

Tasklet’s plans start at $25 a month for casual use and $100 a month for power use, with a $250 tier and any custom amount up to $10,000. Starter at $25 carries 10,000 credits a month, Pro at $100 carries 40,000, and the $250 tier carries 100,000. Billing is at the organization level with unlimited team members and one shared credit pool, so headcount does not move the price. A paid plan can also be scaled by quantity, from $25 to $10,000 a month in $5 steps, each step adding 2,000 credits.

Every request costs credits. Tasklet lists what moves the number: task complexity, the context and data involved, which tools are active, how often an automation runs, and the intelligence level you choose. One-time credits start at $25, at a fixed rate of 320 credits per dollar. They are valid for a year and are spent only after the plan credits are gone. Monthly credits reset each billing cycle and do not roll over, and paid plans add 600 daily bonus credits on days you open the app.

What free means in practice

The free tier is real: a limited set of trial credits, no payment card, and anyone can create an agent and start. Whether it is enough depends on the work. A short research task and a browser session that logs in and fills a form do not cost the same, and credits meter both.

When the trial credits run out, the next task needs a monthly plan or a one-time credit pack. Tasklet’s own guidance is to start on the $25 plan and move up as usage grows. That fits a product whose price moves with work rather than headcount. The free tier settles one question: are these agents useful to you? What a year of them costs is a separate decision, taken at Tasklet’s published credit rate.

Where the work lives and what leaves with you

Everything that makes Tasklet quick to start is also the part you do not hold. The agents, the connections, the vault, the permission rules and the record of what each agent did all live inside Tasklet’s cloud. Tasklet does not publish an open-source version, and its site offers no self-hosted install.

The approval layer and the cloud are one decision for Tasklet. The agent, the policy and the person’s answer all pass through servers Tasklet runs. That is why an agent can stop and ask at all. Move the agent out of that cloud and there is no layer left to hold the answer.

What you can copy is the text. The prompt, the rules about what an agent may touch, the list of tools, the schedule. A careful team can rebuild most of an agent from those four. It takes an afternoon and a text editor. What stays behind is the running system: the sessions, the credentials and the history.

The open-source alternative, and what changes

Kortix keeps those parts on your side of the line. A project is a git repo you own, where the agents, skills, memory and connector settings are files. One isolated sandbox per session. Each session has its own isolated machine and branch. Session work reaches main through a change request. Merge is default-deny for agents. Open source. Read it, fork it, audit it. Self-host is free.

What you take on is real: the install, a model key, and someone who answers when the server is unhappy. What you stop renting is the managed layer, and the repo stays yours if the cloud bill stops.