Screencap

Product Introduction

1. Definition: Screencap is a local-first, open-source (AGPL-3.0) screen recording and workflow intelligence application for macOS. It is a technical tool that captures not just video but also contextual metadata (clicks, keystrokes, window titles, and MCP server context) to create structured datasets from real user workflows.
2. Core Value Proposition: It exists to solve the problem of institutional knowledge loss and inefficient training by automatically documenting how work is actually done on a computer. Its primary value is enabling teams to capture, search, and share real workflows with privacy-by-design, where sensitive applications are blocked at the capture engine level and all data remains on the user's machine by default.

Main Features

1. Privacy-Enforced Local-First Recording: All screen recordings are stored exclusively in ~/.screencap on the user's Mac. The application enforces privacy at the moment of capture, not just in post-processing. It uses a rules-based engine to completely block recording of specified sensitive apps (like password managers and banking software) before any frame is written to disk, and can mask others (like email and chat).
2. On-Device AI Agent for Automated Segmentation & Indexing: An on-device machine learning model processes the raw footage locally. It automatically segments long recordings into discrete, labeled tasks (e.g., "Payroll run," "Ticket triage") and indexes every spoken word and on-screen text. This transforms unstructured video into a searchable database of workflows without sending data to the cloud.
3. MCP (Model Context Protocol) Server Integration: While recording, Screencap queries connected MCP servers (for tools like Linear, Attio, Gusto) to capture a structured snapshot of the specific record or context on screen (e.g., "Payroll run #214 — May"). This metadata is embedded within the recording, allowing viewers to understand and even open the exact live record from the video player.
4. Controlled Sharing with Automated Anonymization: Users can share individual tasks or compile them into onboarding collections. Before any data leaves the machine, a local anonymization process ("scrubbing") runs to redact names, personal identifiable information (PII), and secrets. Users share only the specific, scrubbed content they select.
5. Open Source & Reproducible Builds: The entire codebase—including the capture engine, privacy rules, AI agent, and anonymizer—is publicly available on GitHub under the AGPL-3.0 license. This allows for full security and privacy audits. Application releases are notarized by Apple and are reproducible from source, ensuring verifiable integrity.

Problems Solved

1. Pain Point: The "tribal knowledge" problem, where critical operational workflows exist only in employees' heads or in outdated documentation, leading to training bottlenecks, inconsistent processes, and knowledge loss during turnover.
2. Target Audience: Software & Operations Teams (Engineers, DevOps, QA), Business Operations & Finance Teams (Ops managers, accountants executing complex reconciliations), Customer Support & Success Teams (agents handling intricate troubleshooting), and AI/ML Teams seeking high-quality, real-world interaction datasets for training open computer-use models.
3. Use Cases: Onboarding & Training: New hires can search and watch recordings of how specific, complex tasks are performed. Process Documentation: Automatically creating an audit trail of exact steps for compliance-sensitive workflows like payroll or financial reporting. AI Training Data Curation: Deliberately donating anonymized, real workflow recordings to create public datasets for open-source AI model training, moving beyond synthetic data.

Unique Advantages

1. Differentiation: Unlike cloud-based screen recorders (Loom, Veed) or system-level recorders (QuickTime, OBS), Screencap's local-first and open-source architecture provides unparalleled privacy guarantees. Competitors may encrypt data in transit/at rest, but Screencap prevents capture of sensitive data altogether and keeps the primary copy user-controlled. Its on-device AI segmentation also differs from cloud-dependent analysis services.
2. Key Innovation: The privacy-by-capture model is its core innovation. By integrating the privacy engine directly into the low-level capture pipeline, it technically prevents sensitive pixels from ever being processed or stored, which is more secure than post-recording blurring or deletion. The integration of MCP for contextual awareness also uniquely bridges screen activity with structured application data.

Frequently Asked Questions (FAQ)

1. Is Screencap really private if it records my screen? Yes, Screencap is designed with a privacy-by-design architecture. Sensitive apps are blocked before any video frame is written to your disk, and all processing (AI segmentation, indexing) happens locally on your Mac. Your recordings never leave your computer unless you explicitly choose to share a scrubbed version.
2. How does the AI feature work without sending my data to the cloud? Screencap uses an on-device machine learning model (likely a lightweight, local neural network) to analyze your recordings. All processing for task segmentation, labeling, and speech/text indexing occurs locally on your Mac's hardware (Apple Silicon optimized), ensuring no video or audio data is transmitted for analysis.
3. Can I use Screencap for team training and onboarding? Absolutely. Screencap is built for team knowledge sharing. You can record workflows, which are automatically labeled and indexed for search. You can then share specific task recordings or assemble them into ordered onboarding collections. All shared content is automatically scrubbed of PII before being sent.
4. What does "open source" mean for a screen recorder, and why is it important? Screencap's entire codebase is publicly licensed under AGPL-3.0. This is critical for a tool with deep system access because it allows anyone to audit the code for security, verify its privacy claims, and ensure there are no hidden data collection mechanisms. It builds trust through transparency.
5. What is the difference between the $9/month and the $20/month plan? The $9/month "This Mac only" plan provides unlimited recording and local search, with all data stored solely on one Mac. The $20/month "Personal cloud" plan adds encrypted cloud backup for syncing across your Macs and the ability to share single recordings via link. The Team cloud plan adds a shared, end-to-end encrypted library and collaborative features.