Prostir
Sign in
Try it free

Product roadmap

Progress you can verify. The next risks we are removing.

This roadmap is a product progress record for creators, early customers, partners, and investors. Completed milestones name shipped capabilities; the current milestone names the evidence required before wider access; future work stays conditional on real usage and product readiness.

Build with what works today
  1. 01Built

    Agent and Skill foundations

    Prostir established the two creator products at the center of the platform: an Agent that owns behavior and runtime state, and a versioned Skill that packages a reusable method.

    • Agents combine instructions, supported Knowledge, approved tools, access policy, publication, sessions, and memory under one product identity.
    • Skills package reusable instructions and references with validation, versions, portable ZIP archives, publication, sale, and attachment to Agents.
    • Private-by-default access and exact product ownership were designed into the foundation rather than added after distribution.
  2. 02Built

    Distribution and monetization

    The platform moved from authoring to an owned delivery path: creators can publish a hosted MCP endpoint, choose supported access, and configure eligible payment routes without building the surrounding SaaS backend.

    • Hosted MCP gives each published Agent a stable product endpoint for compatible AI clients, with OAuth-first access and scoped revocable credentials where supported.
    • Website widget and messaging channels reuse the same Agent identity when their separate setup, policy, and provider requirements are satisfied.
    • Eligible Stripe Connect sales and one-time UAH checkout through Rozetka Pay keep the seller and payment provider responsible for the transaction.
  3. 03Built

    Company and Store product foundations

    Prostir expanded from individual AI products into private collaboration and AI-operable commerce while preserving a distinct owner and authority boundary for every product.

    • Company brings private Tasks, Goals, Documents and Files, Knowledge, logical data, approved tools, attached Agents, and a private MCP endpoint into one collaboration product.
    • Store adds catalog, cart, checkout, order, payment, inventory, fulfilment, and an authorized existing Agent around one seller-owned commerce root; Store remains Preview.
    • Flow remains the separate automation product, with durable execution and release acceptance tracked independently from Agent workflows.
  4. 04Built

    Evaluations and controlled Self-Learning

    Product quality can now improve through explicit evidence instead of silent model behavior. Evaluations and Self-Learning are implemented as separate, reviewable modules with exact ownership and revision controls.

    • Agent, Skill, and Company share typed evaluation cases, evaluator profiles, bounded runs, failure signals, and cited improvement proposals.
    • Evaluation output never changes or publishes a product automatically: a person accepts a proposal, and only the exact current draft can be updated.
    • Self-Learning retains only explicit durable preferences, corrections, limits, or rules; shared Company and Store proposals require authorized review and every accepted rule can be revised or rolled back.
    • Company also advanced with Gantt, BPM, lightweight CRM, Operations, and private business-control surfaces for outcome-complete pilots.
  5. 05Now
    We are here

    Remediate, retest, and prove early-access workflows

    The current milestone is to convert product breadth into launch evidence. Prostir has completed its first independent security review; remediation, exact retesting, and bounded customer pilots now gate wider access.

    • Ruslan Siniaiev reviewed the platform in three owner-approved phases: passive reconnaissance, active reconnaissance, and application testing across authentication, payments, access control, and product logic.
    • The audit documented eight vulnerabilities, including one critical issue with direct business impact. Sensitive reproduction and remediation details remain private.
    • Each finding receives an owner, release decision, regression coverage, and retest evidence before it can be considered closed.
    • Early pilots measure one real workflow, safe refusals, human corrections, recovery, time saved, and the product gaps that block a repeatable outcome.
  6. 06NextQ2 2027

    Scale the product paths that earn evidence

    The next horizon is a direction, not a promise. Priority will follow observed use, retained outcomes, security evidence, and the operating cost of supporting each surface well.

    • Deepen high-value channels, integrations, analytics, audit views, and enterprise access controls when real workflows justify them.
    • Broaden Store operations, complete Flow release gates, and turn the most-used Company workflows into stronger end-to-end product paths.
    • Use evaluation failures, security retests, support patterns, and customer outcomes to decide what scales and what stays Preview.

Bring one outcome worth proving

A useful early-access request names the repeated job, its current cost, the information and actions it needs, the person who owns the result, and the boundary that must never be crossed.

Build with what works today
  • Choose Agent, Skill, Company, Store, or Flow by the owner of the work—not by the most fashionable AI label.
  • Define access, payment, delivery surface, approval, rollback, and one measurable pilot outcome before launch.
  • Tell us what success after 30 days would look like and what evidence would make you stop or expand.

Product roadmap

Request early access for a concrete product workflow

Share the outcome, owner, access boundary, and evidence you want to prove with Prostir.

Build with what works today