Skip to main content
Eversince is the workspace where the user’s media and the work on it live. You work in it the way you work in a repository: what earlier sessions made is still here, and a request about it continues that timeline or canvas rather than starting a new one. The user sees each piece on the feed in the Eversince app, on the web, on Mac and Windows or on an iPhone, where they review it, comment on it and edit it. Media comes in on an upload page the user opens from a computer or a phone, and from URLs, online stores or the user’s computer through the desktop app.

​
What an agent can do

  • Finding. Search the library by what is said or shown, down to the moment inside a recording, and find shots that look like a reference. Bring in stock media or an online shop’s products.
  • Understanding. Read transcripts with speakers and shot logs, recognize speakers across recordings, ask a model about a video or a recording, and read PDF, Word and PowerPoint files.
  • Video. Cut many clips out of one recording at once, cut by transcript, and remove silences and filler words. Sync cameras and cut to whoever is talking. Split screens and grids, titles and lower thirds, transitions, keyframes, camera moves, speed ramps, punch-ins, reframing that follows the speaker, cutting to the beat.
  • Images. Layered design across slides for carousels and decks: text, shapes, icons and charts, blend modes, subject and gradient masks. Generative edits to part of a picture, extending a picture to a new aspect ratio.
  • Look. Grade with color wheels, LUTs and masks, 19 effects from film grain to color key, background removal and 4K upscaling on images and video, motion graphics written as code, styled captions.
  • Audio. Remove noise or isolate the voice, level to platform loudness, and duck music under speech. Voiceover, music and sound effects, and stock music with tempo and downbeats.
  • Language. Dub into another language in the speaker’s own voice, or translate the captions, in any of the supported languages.
  • Generation. Images, video and audio from a prompt, a start frame or reference inputs, on the models and at the prices in Models and prices.
  • Checking. Real frames of the composite, color scopes against a reference, and a check for black frames, missing media and loudness.
  • Delivery. Render to mp4, transparent WebM, GIF, mp3, wav, images or PDF. Through the desktop app, hand the cut to Premiere Pro, DaVinci Resolve or Final Cut Pro. Save a timeline or canvas as a template, or share a public link.
  • Organizing. The library works as a file system for media: folders, labels, typed fields, boards and calendars, duplicate finding, and the workspace’s brand kit.
  • Research. Pull posts and ads published on TikTok, Instagram, YouTube, Reddit, and the Meta, TikTok and LinkedIn ad libraries, and keep the ones worth having in the library.
The craft behind these, from rough cuts and grading to UGC ads and trailers, is in skills that manage_skills lists.

​
Three ways to call the same tools

Every tool has one definition and one JSON Schema, available over three interfaces. Work runs on Eversince’s servers: from any host, with nothing open on the user’s side, and a job carries on after the call. The Eversince desktop app for Mac and Windows serves the same three interfaces on the user’s own computer and works on files where they are. See Desktop and cloud.

​
Connecting an agent

An agent sets itself up from one page, eversince.ai/skill.md. The page lists the ways to connect and which one fits the agent’s situation. A user gives their agent that address and nothing else. Connect covers the same steps for people.

​
Calling a tool

GET /api/v1/tools returns every tool with its description and input schema. Over MCP the same list is tools/list; in the CLI it is eversince tools. The API reference is generated from the same definitions.

​
What you work with

  • Library. Every image, video, audio file and document the workspace holds. Each item has an id and an 8-character ref that users copy from the app. Either is accepted wherever an item is named.
  • Timelines. Video documents made of tracks and clips, edited with typed operations that apply as one transaction. A workspace holds any number.
  • Canvases. The same for images: layers on one or more slides.
  • Jobs. Any operation too slow to answer within the call. A job id means the work was dispatched, not that it finished. Follow it with get_jobs, or receive the result on a webhook.
  • Feed and comments. Everything an agent or a person does in the workspace appears on its feed. Comments are the conversation about the work, and they outlast the session that started them.
  • Skills. Saved methods for a kind of job. Eversince’s skills cover the craft of editing, design and generation. A workspace’s skills belong to its users.
The Concepts pages define each of these.

​
Costs

Three balances pay for the work, each in its own unit: AI credits for generation and other model work, cloud processing minutes for everything else that runs on Eversince’s servers, and storage for what the workspace keeps in the cloud. Editing, reading stored results, searching, comments, boards, calendars and public links are free, apart from the few edits that read the picture or the sound; each tool’s description says which. Costs has the full rules, Models and prices and Cloud processing and storage list every price, and estimate_cost quotes any planned call before it runs.

​
Reading these pages as markdown

Every page is available as markdown at its address with .md appended, and llms.txt lists them all.