skip to content

Product roadmap

From data to
action—and back
again.

What we’re building now—and the product milestones that brought us here.

Product timeline

Current and completed product milestones, ordered from newest to oldest.

In progress

The foundations we’re strengthening next to make every workflow more composable, reproducible, and agent-ready.

  1. Modular Pages

    Build richer, more maintainable Pages from reusable local modules—without giving up atomic releases, revision history, or review.

  2. Versioned workspace files

    Give every Action, Model, and supporting file a stable identity with immutable history and safe concurrent editing.

  3. Reproducible Action runs

    Make retries, schedules, and replays execute the exact code, policy, and permissions that originally launched the Run.

  4. Agent-native operations

    Give agents one governed way to inspect Data, edit workspace files, manage Connections, and operate Runs end to end.

AUG2026

Proforma becomes a coherent operating system: govern public software, author with confidence, and learn from every run.

  1. The operating loop comes together

    Connected Data, Pages, Actions, and measurement into one continuous loop from a question to the next best decision.

  2. Pages become real applications

    External agents can build, verify, and publish complete Pages, while viewers can safely use the Actions behind them.

  3. A calmer workspace for long-running work

    Made Plans dependable, introduced persistent work tabs, and clarified the handoff between Chat, Pages, Data, and operations.

  4. Everything is easier to find

    Added product-documentation search, Chat-history search, and safe file organization so past work stays useful.

  5. Stronger agents, safer Page review

    Upgraded Chat, tightened native Prospecting tools, added interaction analytics, and strengthened visual review before publication.

JUL2026

Distribution opens up: teams can publish useful software, share governed Data, and operate through the channels they already use.

  1. Reviewable collaboration

    Made long Chats easy to scan and kept Page feedback attached to the exact version and surface it describes.

  2. Browser work survives interruptions

    Recovered browser queues across cold starts and added lifecycle messages so long-running work hands off cleanly.

  3. Presentation layouts and generated media

    Turned Pages into tailored presentations and let agents create governed image, audio, and video assets inside their work.

  4. Governed access for external agents

    Gave outside agents a bounded view of shared workspace structure without exposing the authenticated product shell.

  5. Visible, dependable Page builds

    Made Page generation, stage handoffs, retries, and failures visible while hardening the public queries underneath.

  6. One coherent Proforma platform

    Aligned Workspace, Data, Pages, Chat, and operations around one production-ready product model.

  7. Shared Data products

    Let teams publish and mount governed datasets, browse shared Models, and query them without copying ownership.

  8. Managed connectors and live market Data

    Made connections discoverable, syncs observable, and live US market feeds reusable across workspaces.

  9. Slack becomes an operating channel

    Turned Slack events into governed agent work and exposed approved Slack capabilities to Actions.

  10. Self-serve signup and meeting memory

    Opened guided signup and added a Google Meet notetaker that turns conversations into durable workspace knowledge.

  11. Operational Data and native Prospecting

    Added app-owned operational tables, managed Postgres mirrors, queryable email, and a governed Prospecting app with native tools.

  12. One inbox, public Pages, and complete history

    Created one attention surface for people and agents, published secure Pages, and made history, freshness, and review visible.

JUN2026

The product gets more legible: trusted definitions, observable work, durable context, and direct routes in and out of the warehouse.

  1. Product analytics and session replay

    Brought product analytics into a first-party dashboard and connected replayed behavior to warehouse events and outcomes.

  2. Trusted metrics and the Data Modeler

    Added a metrics registry, modeling workbench, package loop, trust checks, and Chat references for governed definitions.

  3. Visible Work and live feedback

    Made multi-step work visible, rendered tool activity as it happened, and surfaced Page-query progress and voice input.

  4. Workspace knowledge and Connection control

    Added knowledge search, Notion sync, and one clear place to configure every integration resource.

  5. An analysis API for external agents

    Exposed scoped catalog, query, and analysis tools through MCP so outside agents can work with governed context.

  6. Durable Page context

    Moved long-lived instructions and memory into a visible, editable, revision-aware surface beside each Page.

  7. Faster workbooks and responsive Pages

    Improved workbook first paint, made Pages fit every viewer, and added direct, progress-aware exports for Models and workbooks.

  8. Slack Chat and observable Actions

    Started warehouse-aware Chats from Slack, streamed durable Action logs, and improved mobile history and hosted web research.

MAY2026

Proforma becomes programmable and dependable: a multi-tenant workspace with recoverable automation, safe code, and agent-facing interfaces.

  1. One Workspace tree, with recoverable Plans

    Unified file and Data navigation, made Plans durable workspace views, and strengthened Page save, export, and recovery paths.

  2. Automations you can pause, resume, and recover

    Added explicit controls for interruption and recovery, then brought the same operating surface to the prfm CLI.

  3. An operations control room

    Brought Work, schedules, event triggers, recent Runs, controls, and recovery state into one operator view.

  4. Google sign-in and shared Chats

    Added domain login, workspace-scoped sharing, and clean state isolation when people move between workspaces.

  5. Connections, undo, and safe secret handoff

    Centralized database Connections, made workbook edits reversible, and captured credentials without exposing them to Chat.

  6. Warehouse Models and polished Page export

    Materialized BigQuery Models where teams expect them, added a focused Page-design workbench, and exported production-ready PDFs.

  7. The prfm CLI and Agent API

    Gave coding agents a scoped interface for context, catalog, Data, Pages, and reviewable workspace changes.

  8. Sandboxed Pages and Actions

    Moved authored code into isolated runtimes with capability policies, budgets, tenant-bound caches, and observable limits.

  9. Durable document and market intelligence

    Made PDF extraction persistent and added a live market-research workflow with evaluated agent behavior.

  10. Email operations and managed integrations

    Added an inbox, default webhook ingestion, managed integration controls, and guarded responses from external tools.

  11. Faster, safer Chat and agent Runs

    Reduced latency, bounded tool payloads, strengthened tenant access, and let versioned Actions launch evaluated agent work.

  12. A true multi-tenant product

    Made every route, artifact, query, database, and Run explicitly Workspace-owned, with separate public and authenticated boundaries.

APR2026

The foundation takes shape: a warehouse-native workspace where agents can understand Data, build software, and act on the result.

  1. Files and documents become agent-readable

    Brought CSV, PDF, and XLSX files into the workspace and let agents inspect documents page by page.

  2. Browser work with live visibility

    Let agents operate real browser sessions, show progress live, capture network context, and write results back to the warehouse.

  3. Email, workbook, and voice workflows

    Added email responders, editable workbooks, outbound and conversational voice, and communication tools agents can call.

  4. Team workspaces

    Added membership, invitations, and shared administration so the product could hold real team operations.

  5. Lineage-aware modeling

    Brought freshness, transitive lineage, refresh planning, live previews, and source feeding into both the catalog and agent context.

  6. Durable autonomous runs

    Made long-running work resumable and replayable, with plan–act–reflect steps, budgets, compacted memory, and human requests.

  7. Versioned Pages and operational Chat

    Made Pages editable and restorable, then made Chats branchable, schedulable, lineage-aware, and ready to hand work off.

  8. Warehouse-native product analytics

    Connected event ingestion, a browser SDK, BigQuery delivery, starter Models, Pages, and agent tools into one analytics loop.

  9. BigQuery, the catalog, and Models

    Added schema-aware browsing, raw Sources, SQL-defined Models, materialization, and Actions that could run pipelines.

  10. The full-stack foundation

    Started with a React and Hono product, Postgres identity, DuckDB execution, object storage, AI Chat, and programmable Actions.