Privacy policy
Last updated: August 2, 2026.
This policy covers Agentissimo, the hosted backend, and the local daemon. Agentissimo keeps code execution local. The hosted backend handles account auth and coordination; it does not execute agents or hold model keys.
What stays local
Section titled “What stays local”The local daemon runs on your machine and binds loopback only. Agent CLIs, workspace files, repositories, prompts, model responses, model API keys, and agent execution stay on your machine.
Agentissimo does not send WorkOS or the hosted backend:
- Source code.
- Repository contents.
- File paths or repository names.
- Prompts or model responses.
- Model provider API keys.
The daemon keeps some analytics data in its own local store so the dashboard can report over a window — for example the Attention tab’s interruption record, which stores the workspace, the session, why the agent needed you, and when the interruption opened and closed. It records no message text. This data stays in the daemon’s data directory on your machine and is never sent anywhere.
Account data
Section titled “Account data”Agentissimo requires an account. Account identity is provided by WorkOS AuthKit. WorkOS may process the account information needed for signup, login, session management, MFA, organization membership, social login, and Enterprise SSO when configured.
Account data may include:
- Email address.
- Name and profile fields returned by the identity provider.
- User id, session id, and organization id.
- Organization membership.
A WorkOS Organization maps to an Agentissimo tenant. Product entitlement is stored by Agentissimo as tenant plan data; WorkOS is not the source of truth for plan entitlement.
Backend data
Section titled “Backend data”The hosted backend stores the minimum data needed to authenticate users, scope tenants, coordinate runners, and route collaboration traffic.
Backend records may include:
- Tenant ids and plan tier.
- Session and runner ids.
- Timestamps, status values, counters, and protocol metadata.
- Opaque handles needed to route or manage backend-owned records.
The backend must not persist prompts, model responses, repository content, or model-key material. Brokered runner traffic is treated as an opaque envelope.
Telemetry
Section titled “Telemetry”Agentissimo collects no telemetry by default. When you opt in via Settings → Skill Usage, aggregated skill-usage counts (day, catalog, skill, harness, count) are synced to the backend to power marketplace rankings. No model token counts, prompt content, repository paths, or key material are included. You can disable sync and delete your records at any time from the same settings page.
Feedback and issue reports
Section titled “Feedback and issue reports”Send Feedback is user-initiated. You pick a category (Issue, Feature Request, General Feedback, or Other) and Agentissimo emails the message you wrote, any images you attach, and, for an Issue while the diagnostics option is on, a diagnostics bundle. The form lists exactly what that bundle carries: app version, operating system and browser, language and timezone, window size and the page you are on, and recent in-app error messages. It never includes source code, repository contents or names, file paths, prompts, model responses, or keys. Your reply-to address is the account you are signed in with, and a copy is emailed to you only when you ask for one. Nothing is sent until you submit, and no report is gathered in the background.
Skill security scanning
Section titled “Skill security scanning”Curated marketplace skills carry a “Scanned by Agentissimo” attestation. The scan runs in a curation pipeline separate from the backend service: it fetches the public catalog content from GitHub, checks it against a static ruleset plus an AI review, and discards the content. Only the verdict, finding locations, and a content hash are stored — never the skill’s file bytes. The AI review runs under a single-purpose key held by the pipeline; the backend holds no model key. Skills you upload are checked with the static ruleset only, from the bytes you already submitted; Agentissimo runs no AI review over your content. The optional per-skill Advanced scan in the app runs through your own configured model, locally — its results are never uploaded.
Sub-processors
Section titled “Sub-processors”WorkOS is used for identity and authentication. WorkOS receives identity data needed for AuthKit login, refresh, logout, JWKS verification, organization membership, and Enterprise SSO when configured.
WorkOS must not receive session content, code, prompts, model responses, runner envelopes, or repository data.
Retention, deletion, and export
Section titled “Retention, deletion, and export”Agentissimo keeps account and backend records only as long as needed to operate the product, meet security obligations, and honor customer agreements.
Deletion and export requests are tenant-scoped. Deleting a tenant must cascade to that tenant’s backend-owned records.
Security
Section titled “Security”Agentissimo separates the hosted backend from the local daemon. The backend verifies authenticated sessions, scopes requests by tenant, and does not link into the shipped local daemon binary. Secrets are injected at runtime and must not be committed into source or baked into images.
Changes
Section titled “Changes”Agentissimo may update this policy as the backend and account surfaces change. Material changes should be reflected in this page before the changed behavior operates with customer data.