Startup name: Lida
Tagline: An expert beside you, on your screen.
Elevator Pitch: Lida is a Windows desktop AI companion that sees your screen and talks with you. In Guide mode it circles the next thing to click, in Teach mode it explains what’s on screen on a board, and in Agent mode it does the task while you watch, asking before anything risky. Ctrl+Alt+Space stops it at any time.
Connect Gmail, Calendar, Notion, GitHub, Jira or any MCP server. Currently in private preview; join early access at https://trylida.com/early-access.
Target Market: Windows users who want on-screen guidance, learning, or task automation
How will you make money?: Free tier; paid Pro ($20/mo) and Max ($50/mo) plans planned.
How much capital have you raised?: null
Website: https://trylida.com
City/Country: India
Lida is a Windows desktop companion that uses voice and screen context to help people navigate software, learn workflows, or complete tasks while they watch. Its premise is notable because it puts assistance directly over the application a user is already using, rather than sending them to search results, videos, or a separate chat window. (Source: https://trylida.com/)
Many software-help experiences require users to interrupt their work, describe their context manually, and translate general instructions back into the specific interface on their screen. Lida is designed around that friction: it says it can inspect the active app, window, and visible controls during a session, then point users to the relevant next step. (Source: https://trylida.com/about)
The company’s published use cases suggest an initial audience of people learning or troubleshooting desktop software, including Excel and Blender users, as well as Windows beginners and people seeking home tech support. (Source: https://trylida.com/use-cases)
Lida is currently presented as a Windows desktop app in private preview, with access distributed through invitations. The company says the app can interpret what is on screen during a user-started session and provide assistance through spoken responses, on-screen visual cues, and captions. (Source: https://trylida.com/product/desktop)
The product has three operating modes. In Guide mode, Lida moves a cursor and highlights the control a user should select; in Teach mode, it explains a concept with a narrated, captioned visual lesson; and in Agent mode, it can click and type on the user’s behalf while the user watches. (Source: https://trylida.com/product/features)
The agent is positioned with explicit human-control safeguards: Lida says it pauses if the user touches the mouse or keyboard, requires approval before risky actions, and will not enter passwords or card numbers, transfer money, or change security settings. (Source: https://trylida.com/security)
Its privacy approach is central to the product proposition. Lida states that screen content is processed during active sessions to provide help, is not saved as images, and can be blanked when password managers, banking apps, or user-blocklisted apps are in front. (Source: https://trylida.com/security)
Lida appears to be pursuing a credit-based subscription model in which one credit represents roughly one minute of active assistance. Its published, but explicitly provisional, plans include a free tier with 30 monthly credits, a $20-per-month Pro tier with 600 credits, and a $50-per-month Max tier with 1,800 credits. (Source: https://trylida.com/pricing)
The company also lists provisional 300-credit packs for $10, while noting that checkout has not launched, users cannot currently pay through the site, and final pricing may change after it measures real operating costs. (Source: https://trylida.com/pricing)
Publicly visible traction is limited to the private-preview status and invitation-based rollout; the reviewed materials do not provide citable customer, revenue, funding, or team metrics. (Source: https://trylida.com/about)
Lida’s strongest positioning is its attempt to combine contextual software assistance, instruction, and supervised computer control in one interface. The product’s emphasis on visible state, interruption controls, approval gates, and restricted actions is sensible for an agent that operates on a user’s desktop.
The main open questions are execution quality across the long tail of Windows applications, reliability when visual context is ambiguous, and whether per-minute pricing can remain understandable and attractive as model-serving costs evolve. Its private-preview stage means the commercial and retention case is still unproven.
Note: Information based on publicly available sources at the time of writing, and summarized by AI.
AUCTIGO Global - B2B marketplace for electronics offers, buying requests and auctions
Kenerate AI - An AI creative studio for images, video, voiceovers and music.