AI tool listing.
Acti is a mobile AI assistant built into the keyboard layer rather than into a separate chat application. Its central interaction is simple: type an intent in any text field, hold the Acti Bar, inspect the proposed result or action, then apply it without leaving the host app.12 That makes the product structurally different from most AI keyboards, which primarily improve the text being typed through autocorrect, rewriting, translation or dictation. Acti tries to turn the keyboard into an execution surface.
The current v2 product is available on iOS and Android and is designed to work wherever a third-party keyboard can appear. The official examples include finding restaurants or sports information, retrieving Notion documents or LinkedIn profiles, creating Google Meet links, using Calendar actions and running custom workflows.13 The useful idea is not that any one of those tasks is technically unique; it is that the user does not have to leave the message, email, note or other text field where the task originated.
The Acti Bar is the product’s general-purpose trigger. Pressing behaves as normal keyboard input; holding invokes the agent. Acti says the agent previews the reply, link, summary or other result before the user applies it, rather than silently executing every generated action.2 That confirmation layer matters because agentic mistakes are more consequential than a bad autocomplete suggestion: an incorrect generated sentence is annoying, while an incorrect calendar event, email or connected-tool action can affect another system.
Acti’s second major layer is Skills. A Skill packages a reusable action and can be bound to a key. Skill Builder lets a user describe the desired workflow in natural language; Acti assembles the prompt, tools and output without requiring code.4 Skills can remain private or be published to the Skill Hub, where other users can install or clone them. Acti says more than 1,000 Skills had already been created around launch and now presents the Hub as an open programmable-keyboard ecosystem.56
The extensibility is meaningful because the keyboard itself is only the surface. Acti’s current marketing says it can connect to more than 100 tools and APIs, while the product pages describe integrations such as Gmail, Slack, Notion, Google Calendar and other OAuth-based services.17 A well-designed Skill can therefore combine retrieval, generation and a connected action. This is closer to lightweight workflow automation than to predictive typing.
There is still an important ceiling. Acti is not a replacement for a full agent workspace when the task requires long research, complex visual inspection, substantial file manipulation or prolonged multi-step supervision. Its advantage is immediacy inside mobile text workflows. Acti’s own comparison material frames dedicated assistants as better suited to long-form, open-ended conversations and research.8 The keyboard form factor is strongest for short actions that begin or end in a text field.
The product is also young enough that the base keyboard itself remains part of the evaluation. Current app-store reviews praise speed, autocorrect and the concept, but request more language support, keyboard sizing/customization, wider app compatibility and better iPad behavior; some launch-period iOS reviews described basic keyboard failures that later updates may already have addressed.93 Since the keyboard replaces a core daily input surface, ordinary typing quality matters as much as the AI layer.
Acti is currently free. The Terms of Service state explicitly that Acti does not sell subscriptions, paid plans, paid Credits or in-app purchases and does not accept payment for the Service. AI-powered Skill runs are controlled instead through a free internal Credits allowance with daily and rolling limits; Credits have no monetary value and cannot be purchased.7
The exact free allowance is not durably quantified in the current public Terms. One current Google Play reviewer reports a limit of 10 free AI uses per day, but that is a user observation rather than a stable first-party entitlement and Acti reserves the right to adjust Credit allowances.37 The product should therefore be treated as free with usage quotas; the user-reported “10 per day” figure is not a guaranteed current entitlement.
There is a documentation mismatch worth preserving. Google Play currently labels the Android app as offering “in-app purchases,” while Acti’s contractual Terms say there are no paid features or purchasable Credits at present and that any future paid features would receive additional terms when introduced.37 The Terms are the more precise current billing statement, but the store metadata should be reconciled by the vendor because it can confuse users evaluating cost.
TechCrunch reported at launch that Acti expected eventually to monetize through subscriptions offering more advanced models, higher usage limits and other premium features.6 That was a stated future business direction, not evidence that such subscriptions exist today. The current Terms supersede that launch expectation for present pricing.
The free positioning is attractive for experimentation because the keyboard, Skill Builder and Skill Hub can be tested without a paid commitment.45 The trade-off is predictability: a free Credit system whose limits can change is less suitable for a workflow that must execute a fixed number of business actions every day. Until paid service levels or stable quota documentation exist, heavy users should treat capacity as a product variable rather than a contractual allowance.
Acti has more real-world evidence than many newly launched AI tools, but it is still an early-stage product. Google Play currently shows more than 100,000 downloads, roughly 194 reviews and an aggregate around 4.5/5.3 The US App Store currently shows 23 ratings at 4.1/5.9 Both are exact-product, current app-level signals and are retained as external rating inputs; the Android signal deserves substantially more statistical weight because its sample is larger.
The qualitative reviews are broadly encouraging but also reveal the maturity gap. Android users praise the idea, typing experience and Skill Keys while requesting more customization, languages and features. iOS reviews similarly praise speed and the concept but identify iPad/landscape support, compatibility and launch bugs.39 These are useful because they concern the actual keyboard rather than an upstream model benchmark, but neither store yet provides the depth of evidence available for mature mainstream keyboards.
Product Hunt adds traction context rather than a defensible rating. Acti launched first for its day, week and month and currently shows 5.0/5 from only three reviews with roughly 3,200 followers.10 The tiny rating sample is excluded from numeric rating inputs. The comments are nevertheless directionally consistent with the product thesis: users value reducing app switching and retrieving or executing something from the keyboard.
TechCrunch provides the strongest independent launch coverage currently available. It confirmed the iOS/Android launch, the Gemini-based AI stack described by the company, more than 1,000 early Skills, the local-first design claim and a $5.3 million seed round led by BITKRAFT Ventures.6 Digital Trends independently described the interaction as an agentic keyboard that performs tasks inside the apps users already use.11 These sources establish product identity and external interest, but they do not constitute long-term reliability testing.
Privacy is unusually important because third-party keyboards are inherently sensitive. Acti’s current Privacy Policy is specific: ordinary text typed through the keyboard is not collected, stored or transmitted to Acti’s servers for storage or training. The keyboard reads surrounding text locally for standard typing features; text leaves the device only when the user explicitly triggers a Skill, and then only the input that Skill is configured to read.12
That does not mean all AI processing is local. When a Skill runs, its explicit input and instructions can be sent to Acti services and third-party AI model providers to generate the result. Acti says those AI providers are contractually prohibited from using the content to train their general-purpose models.12 This is a materially narrower claim than “nothing leaves the device,” and users should understand the difference.
Connected accounts receive another layer of protection in the published policy. Acti says it calls Gmail, Slack, Notion and other OAuth-connected tools only when a user explicitly triggers a Skill that requires them; it does not silently mirror or synchronize those accounts, does not use connected-tool data for AI training, and processes fetched data transiently rather than keeping a stored copy, aside from limited troubleshooting metadata.12 For Google API data specifically, Acti states compliance with Google’s Limited Use requirements and says Google user data is not sent to AI providers.12
Those are strong first-party commitments, but they are still vendor assertions rather than a public independent security audit. A keyboard with network access and OAuth connections has a higher trust burden than a standalone note-taking app. Organizations with mobile-device management or strict data-loss-prevention policies may block third-party keyboards regardless of Acti’s architecture.
The Terms appropriately emphasize human confirmation. Users are responsible for reviewing AI output before it is sent or saved through a connected tool, and Acti warns that the AI can misread intent, choose a wrong recipient or include incorrect content.7 This is consistent with the Acti Bar’s preview-before-apply design, but it also defines the product correctly: Acti accelerates actions; it does not eliminate the need to supervise them.
Acti has exact app-store ratings, independent launch coverage and detailed first-party privacy/legal documentation, but it still lacks mature enterprise evidence, long-term reliability data and a large independent review corpus.
Acti is best suited to mobile power users whose work is fragmented across messaging, calendar, notes, search and lightweight business apps. If a user repeatedly leaves a conversation to find a link, retrieve a document, translate something or create a meeting, the keyboard layer removes a real amount of friction.16
It is also a good fit for users who like building personal micro-automations but do not want to learn an automation language. Skill Builder turns a natural-language description into a reusable Skill and lets the user bind it to a key or publish it to the community.4 That makes Acti unusually approachable compared with agent frameworks that begin with APIs, workflow graphs or code.
The strongest use cases are short, context-adjacent actions: retrieve a result, create a link, draft or transform text, invoke a connected service and return something to the current field. Users who primarily need deep research, long documents, extensive file work or complex autonomous workflows will still need a dedicated assistant or agent environment.
Acti is a weaker fit for users who demand years of keyboard polish, broad language/layout support or perfect tablet compatibility. Current store feedback still asks for customization, additional languages and iPad improvements.93 A third-party keyboard must earn the right to replace the system keyboard every time the user types, so small ergonomic defects have disproportionate impact.
It is also a poor fit for organizations that forbid third-party keyboards or OAuth-based agent actions on managed devices. Even with Acti’s local-first typing design, explicitly triggered Skills can transmit selected text to Acti and AI providers, and connected Skills can act inside external accounts.127 That may be acceptable for personal productivity but incompatible with some corporate policies.
Products matched by shared tasks, audience, workflow, product type and capabilities.