Melete is an AI workspace for teams, with the core idea of letting AI agents work within the tools a team already uses, with human approval before important actions.
Purpose Melete is designed to handle recurring handoffs and paperwork between teams or within a profession. Examples include weekly reports, customer handoffs, release briefs, and research, work that often depends on context scattered across multiple tools. Users only need to describe the outcome they want, and the agent will gather context from connected tools, carry out the work, and hand the result to the appropriate person for review before it is sent.
Key capabilities
- Cross-tool work: Agents work in Slack, Notion, Linear, and GitHub through the user's connected accounts, using only the access permissions the user has chosen. Google Calendar, Microsoft Outlook, Microsoft Teams, and others are supported.
- Projects and approvals: Each project contains requests, the sources used by the agent, tasks in progress, and items awaiting human decisions, so everyone who joins can see the progress of the work.
- Delivery from request to human review: The agent gathers context, carries out the work, and hands the result to the appropriate person before it is sent.
- Readable and correctable memory: Melete remembers decisions, terminology, and preferences, and each memory has a scope (company-wide, a specific project team, or just yourself).
- Isolated task environments: Each task has its own browser, files, and code tools, isolated from personal computers and limited to connected tools.
- Background work and schedules: You can schedule Monday summaries or daily checks, with notifications when there are changes or decisions needed.
- Verifiable deliverables: Briefs, plans, and memos include sources and open questions so readers can verify them, and they can be shared or exported.
- Actions with consequences await approval: Users can choose which actions require approval, reviewers can see what is about to happen, and decisions are recorded.
- No repeating completed steps: Resuming an interrupted task will not repeat completed steps; actions that cannot be confirmed are marked as unconfirmed and needing investigation rather than retried.
- Complete records: You can see what the agent read, what it proposed, who approved it, and what was changed.
- Model choice: You can use OpenAI, Anthropic, or Google models, or run your own models; when switching, projects, memory, and documents remain unchanged.
- Personal assistant: Individuals can use Melete without a company workspace, and personal tasks and memories are kept separate from company work.
How to use it Start with a single workflow: choose a repetitive one that already involves multiple people and multiple toolsSexual intercourse, first prove its effectiveness before connecting more. Connect only the necessary channels, folders, and topics, starting with read-only and drafting permissions. Track the corrections people make, whether results are reused, and where review slows progress, and expand access permissions only after the workflow stabilizes. Teams can join the company waiting list, and the founder will reach out to set up the first workflow.
Applicable scenarios Repetitive work divided by profession, such as teachers (paper feedback drafts, parent inbox), legal teams (deadline calendars, case file indexing, redline review), real estate agents (pre-listing comparisons, post-viewing follow-ups), accountants (collection documents, month-end anomalies), recruiting (interview loops, screening, debrief preparation), consultants and agencies (client status reports, RFP first drafts), small business owners (supplier invoices, reviews, scheduling, renewals), researchers and scholars (literature matrices, peer review responses), nonprofits (grant reports, year-end donation letters), property managers (maintenance triage, lease renewals).
Limitations and notes The demos shown on the website are for testing only, and the appearance and functionality of the real product will differ. Mentioned pricing, integrations, and limitations need to be checked against requirements, but no specific pricing information is given. On security, agents work in isolated environments, but specific compliance or data retention measures are not yet provided. Personal use and company workspaces are separate, but specific account management details are not yet provided.
