Cofia
Privacy-first AI automation platform whose desktop agent learns how a team works from system events and anonymized network traffic, then proactively generates executable automations for human review, with no prompts or manual setup.
Cofia is a New York City company (YC Winter 2026, founded 2026 by Paola Martinez and Moses Wayne, ex-Duolingo and Brilliant.org) building a privacy-first AI automation platform that learns how a team actually works and proactively generates automations without prompts or manual setup. Rather than asking users what to automate, a desktop agent observes system events and anonymized network traffic, mines repeatable workflow patterns, and once a pattern is stable and confident, uses LLMs to synthesize an executable tool-calling agent to replicate it. Generated agents are presented for human review before activation, credentials are handled by watching OAuth flows the user already completed, and the company emphasizes that it never screen records and never captures sensitive data like banking or passwords, with enterprise-grade encryption and user-visible transparency into what is processed. Cofia targets internal operations at high-growth technology companies and reports early ARR. As a very early-stage product, much of its agentic depth is described at a roadmap and architecture level rather than documented as shipped, released capabilities.
Vendor details
Canonical URL
https://www.cofia.ai
Category
Enterprise operations agent
Funding status
Private, pre-seed. Founded 2026 in New York City by Paola Martinez and Moses Wayne; Y Combinator Winter 2026 batch, backed by Y Combinator and TMV, with USV noted as an early backer.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Observes work through system events and anonymized network traffic rather than screen recording, and handles credentials by watching OAuth flows the user has already completed; specific third-party connectors are not enumerated publicly.
Sources & related URLs
Related / legacy domains
Capability coverage
5.5 / 14 capabilities · 39%
| Integrations & Tool CallingObserves via system events and anonymized network traffic and reuses existing OAuth credentials; specific connectors are not publicly enumerated (cofia.ai, startuphub.ai). | Partial |
|---|---|
| Workflow OrchestrationChains observed steps into multi-step automations, but pattern-to-agent synthesis is early and not documented as broad orchestration (cofia.ai, startuphub.ai). | Partial |
| Knowledge Grounding & RAGMines a workflow pattern corpus and federated patterns across users to ground automations; not a document RAG system (zeitgeist.bot, startuphub.ai). | Partial |
| Human Oversight & GuardrailsGenerated agents are presented for human review before activation (startuphub.ai). | Full |
| Security, Identity & GovernanceEnterprise-grade encryption, excludes banking and password apps, and user-visible processing; no named certifications like SOC 2 yet (cofia.ai). | Partial |
| Observability & AuditabilityUsers can see what is being captured and a user-accessible audit trail exists; no per-run agent trace documented (cofia.ai, startuphub.ai). | Partial |
| Memory & State PersistenceBuilds a persistent workflow pattern corpus per user and across the user base; no first-class agent memory framework documented (startuphub.ai, zeitgeist.bot). | Partial |
| Deployment & Data ResidencyDesktop agent plus cloud processing; no self-host or residency options documented (cofia.ai). | Partial |
| Prebuilt Agents / Templates / PacksAutomations are discovered per user rather than shipped as prebuilt templates or packs (cofia.ai, startuphub.ai). | Unable to verify |
| Triggers & Channel CoverageProactive, event-driven detection of repeated patterns; channel coverage is the desktop environment rather than broad triggers (cofia.ai). | Partial |
| Model Flexibility & RoutingNo user-facing model choice is documented (cofia.ai). | Unable to verify |
| APIs / SDKs / MCP ExtensibilityNo public API, SDK, or MCP is documented (cofia.ai). | Unable to verify |
| Testing, Debugging & OptimizationNo agent testing or evaluation tooling is documented (cofia.ai). | Unable to verify |
| Browser / Computer-useA desktop agent observes and can act within the user's software environment, but computer use is framed around observation and OAuth rather than full autonomous computer control (startuphub.ai, cofia.ai). | Partial |
Pricing
No public pricing; request early access, with pricing likely custom based on organizational needs.
Not published; early access.
Cost watchouts
As a desktop agent that observes work and runs cloud automations, cost will likely track seats and automation volume, and enterprise trust and security review add adoption overhead; none of this is published.
Variable cost rationale
Cost likely scales with seats and the volume of generated automations, but no rates are published to anchor it.
Overage / add-ons
Not published.
Sales call required
Yes — required for paid access
Free / trial
No free tier or trial documented; access is via early-access request.
Lowest paid plan
Not published.
Key ambiguities
No pricing model, tiers, or rates are published for this brand-new product.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Akro AI — On premise operational intelligence that automates document heavy…
- Algebra AI — Delivery led provider that builds and runs human governed AI agents…
- Alloy.ai — Commerce intelligence system for consumer brands that unifies four…
- Apaleo — Open, API first hospitality property management platform pivoting…
- Apprentice.io — AI native manufacturing platform for pharma, biotech, and cell and…