One entity. It answers wherever you talk — desktop, voice, Telegram, Slack, Teams, email — and it makes the systems your work runs on legible to it: ask, and it helps you connect them in chat when you need them.
Public preview. Cross-platform CI passed;
field soak and partner clean-install evidence were waived for
this preview and are not complete. Windows may show SmartScreen because its installer
is not Authenticode-signed; Linux .deb updates are
manual.
Fastest path: Create the account in your browser first, then open Studio on your computer, sign in, and ask your first real task in minutes. No sales call and no integration setup required to start. Already have an account? Sign in.
What's on my calendar tomorrow? Draft a prep note for the 10am.
Pull the action items out of this project folder and draft tomorrow's priority list.
You can start without connecting a cloud service — after granting any normal OS permissions.
Choose Studio if you want to start using Evidloom yourself on your computer right away.
Choose design partner if you want help shaping one real team workflow across multiple systems with founder involvement.
Running a small AI-product team with repeated manual work across tools and chat? Send a short design-partner note — we usually reply within two business days. No sales call required for individual Studio use.
Things you say to it
Start in chat. Connect systems conversationally, then use Settings for account, permissions, and troubleshooting. The same entity follows you across every channel you've paired.
Same entity on every channel — same memory, same skills, same ongoing work. Ask on Slack what you started on Telegram; it was there.
The right context, when it matters
Evidloom brings forward the relevant instructions, source material, and tool for the task in front of it. It keeps the working context focused. Irreversible deletion, moving money, and sending secrets or personal data externally require your explicit confirmation.
Artwork AI-generated for Evidloom using Reve.
One entity
You don't get an app with integrations. You get one agent that inhabits the channels you already use, and that can reach the systems your work runs on — files and screen locally, your company's tools through connectors and MCP. Pair a channel once; the entity is just there.
Telegram, Slack, Teams, and email-in ship with owner-only pairing — messages from anyone but you never reach the agent. Email-in runs through a connected Microsoft account. WhatsApp is not supported. Further HTTP-based channels can be added from a declarative adapter specification. When the agent hits a missing capability, it can write the tool it needs. Drafts pass deterministic schema, redaction, injection, and coherence checks, and an authored tool only runs after you approve the source you were shown.
How it works
You talk. It acts. When Studio is running, a scheduled cycle at 3am local time asks the agent to consolidate recent work and check configured tools and channels. Repairs are best-effort, and unresolved failures are reported.
Type to the agent in a dedicated desktop window. It reads files you share with it, edits them, runs commands you approve, and remembers the context across turns. Browser automation is built in for research and form-filling.
Studio combines computer access, voice, and reviewable tool authoring — eyes (screen), hands (mouse, keyboard, files), and a conversational interface.
Today the agent acts inside your files, browser, terminal, screen, calendar, mail, Telegram, and anything you reach via MCP — GitHub, Notion, your warehouse, your password manager. When it hits a workflow it doesn't have, it can write a tool for it, which runs once you approve the source.
When Studio is running, a scheduled cycle at 3am local time asks the agent to consolidate recent work and check configured tools and channels. Repairs are best-effort, and unresolved failures are reported.
Voice mode
Click the voice button and the chat dissolves into a circular field of pulsing lights that breathe with the conversation. Ask in plain English; the agent replies in a natural voice. Voice and chat are the same entity — same memory, same skills, same tools, same MCP servers, same custom workflows you've taught it. The only thing that changes is whether you're typing or speaking. End the session and you're back in chat with the full transcript so nothing's lost.
Pricing
Studio includes $40 of managed model usage, with prepaid top-ups available. BYOK costs $10/month plus charges from your model provider.
Evidloom does not sell personal data or use your content to train Evidloom models. Your selected model provider handles prompts under its own API terms and data controls.
Evidloom does not sell personal data or use your content to train Evidloom models. Your selected model provider handles prompts under its own API terms and data controls.
Both tiers support bundled skills and configured MCP servers; the agent can author additional tools, which run after you approve them.
Trust
Evidloom Studio is local-first. Conversations live in plain files on your disk so you can back them up, search them, or delete them without our help. We meter LLM usage through a hosted proxy because that's what makes the included-usage + flat-rate top-up pricing work — and we say plainly what stays local, what passes through our service, and what never becomes a cloud copy of your machine.
Every conversation, project, and skill the agent learns. Most locally configured tool credentials. Your local files. The agent runtime itself. BYOK provider keys are transmitted over HTTPS and stored in hosted encrypted infrastructure so the proxy can use them. We do not upload a replica of your Mac or PC into our cloud.
Prompt and response text passes through our hosted proxy while we route it to your chosen model provider. We meter usage and keep billing, account, and lightweight runtime state — not a mirror of your local files.
A one-click Repair install in Settings to reset the runtime. A one-click skill-share toggle. A clear billing dashboard. Delete local history whenever you want. Your data stays yours; we keep the service running.