Think-Act puts writing tools, OCR, document analysis, screen recording, and a Desktop Agent in one Windows app. Use a tool on its own, or ask the agent to handle a clearly defined task that needs several supported steps.
Think-Act Workspace Screenshot
Recommended: 1920 × 1080
Most desktop tasks start with something you already have: a report, screenshot, scan, selected passage, or recording. Add that material to Think-Act instead of describing it again from memory.
Using a separate app for every step means more copying, pasting, and switching windows. Think-Act keeps the source, the tools used, and the result together so you can check the work in one place.
Bring in the source, use the tool you need, and keep the result and review connected to the same task.
Context & Files Screenshot
Recommended: 1440 × 900
Give Think-Act the actual file, selected text, screenshot, scan, or recording behind the task. OCR, transcription, and document analysis can then pull out the details you need.
Writing, OCR, document analysis, and recording all work on their own. Use the Desktop Agent when a task needs more than one tool, an enabled connector, or an approval before an action runs.
Desktop Agent
Add the source, describe the result, and choose which tools are available. The agent works through supported steps, pauses when approval is required, and records activity you can review.
Each run uses only
Writing Assistant
Use the Writing Assistant to rewrite, clarify, translate, summarize, or draft from selected text and attached context.
Writing AssistantOCR & document intelligence
Use OCR and document intelligence to extract text or fields, inspect PDFs, and pass verified information into the next step.
Document intelligenceScreen Recorder
Use the Screen Recorder to capture a process and turn its transcript into a walkthrough, SOP, bug report, or future task context.
Screen RecorderSkills & connected tools
Save instructions that worked as a local Skill. Add external tools through trusted MCP connectors only when the task needs them.
Skills & integrationsNot every job needs an agent run. Rewrite selected text, extract a scanned field, inspect a document, or record a process directly. If the job grows, that output can become verified context for a larger workflow.
Writing Assistant Screenshot
Recommended: 1600 × 1000
OCR Screenshot
Recommended: 1400 × 900
Screen Recorder Screenshot
Recommended: 1400 × 900
Here is one example: review a scanned supplier invoice against a purchase order and prepare a list of differences. The agent can use only the source material, built-in tools, and enabled connectors available to the task.
Execution Log Screenshot
Recommended: 1200 × 800
Context
Provide the scanned invoice, matching purchase order, and the fields or exceptions you need checked.
Understanding
OCR and document intelligence can read the supplier, invoice number, totals, due date, and line items.
Example: flag differences in totals, tax, dates, or line items for review.
Execution
The Desktop Agent can use the document tools and enabled connectors needed to compare the records and list the differences.
Human decision
If the workflow would update a record or send the notes, that action can wait for your approval.
Review
Check extracted values and flagged differences against the invoice and purchase order before relying on the result.
Approval Controls Screenshot
Recommended: 1440 × 900
Choose which connectors a task can use and whether each supported action can run, needs your approval, or stays blocked. This puts the check before the action instead of after it.
Enable only what the task needs
Connectors not available to a run cannot be used by it.
Allow, ask, or block per action
Set each supported action type according to its impact.
Review what the agent did
Inspect tool calls, results, and observations after a run.
Unavailable connectors cannot be used
Their tools return only after the connection works again.
Start with one real task. Check the result, adjust the instructions, and save the steps that worked as a local Skill you can use again with new material.
Run
Start with a task where you can judge whether the result is correct.
Verify
Check the result against the source and adjust any instructions that did not work.
Capture
Save the useful instructions, requirements, checks, and expected result as a local Skill.
Reuse
Use the same instructions with new material. Connector access is still controlled separately.
Think-Act works best for specific tasks that involve moving information between documents, screenshots, recordings, writing, and approved business tools.
Summarize reports, improve correspondence, prepare action lists, and create clearer notes for the next person.
Office workflowsExtract errors from screenshots, turn recordings into reproduction steps, prepare issues, and preserve useful debugging procedures.
Developer workflowsExtract invoice and statement data, compare totals, prepare tables, and keep original sources close to review.
Finance workflowsWork through scanned agreements and policy text, then prepare clause notes, key dates, and open questions for qualified review.
Legal workflowsDocument procedures, prepare onboarding material, create SOPs, and preserve reviewed process knowledge as reusable skills.
Operations workflowsExtract text from images, simplify documents, improve everyday writing, translate content, and document recurring personal tasks.
Personal useAnswers about source files, connectors, approvals, and where processing happens. For more, visit the full FAQ page.
No. The Desktop Agent can use only its built-in tools and tools provided by enabled connectors. If an application does not have a supported connector, the agent cannot use it. Review integrations and tool access.
Browser AI usually starts with a chat, so you have to bring the source into it. Think-Act can start with desktop material you choose—selected text, a screenshot, document, file, or recording. Its Desktop Agent can then use enabled tools for a multi-step task. Actions in other systems still need a supported connector and the required approval.
No. Writing, OCR, document intelligence, and screen recording work without a connector. You need a connector only when the task must use another service or local tool. See how integrations work.
It depends on your approval settings. Supported actions can run automatically, pause for your approval, or stay blocked. Think-Act cannot use a connector or tool that is not enabled for the task. Review the security and approval model before allowing important actions to run automatically.
Not necessarily. Think-Act runs as a desktop application, but processing location depends on the model endpoint, connectors, and workflow you configure. A local MCP server may also make network calls according to its own code. Review the data-flow guidance before using sensitive content.
Rewrite a message, pull data from a scanned document, record a process, or give the Desktop Agent a specific task. Check the result, then decide whether you want to use that workflow again.