# ParleHub > ParleHub is a team-based shared workspace for AI conversations, built for organizations rather than solo users. Projects hold shared AI threads every member can see live, each with its own budget, the customer's own expense code, and a token ledger that exports to CSV. Teams either bring their own model provider keys or use models ParleHub hosts, and can run both at once. Sign up at https://app.parlehub.com. ParleHub is made by ParleHub LLC (North Carolina, USA). This file describes the marketing site at https://parlehub.com. The product itself lives at https://app.parlehub.com. ## What ParleHub does - **Shared project workspaces.** A project contains AI conversation threads visible to every member in real time, not per-person silos. Live presence shows who is viewing, typing, or waiting on a response; a teammate's in-flight message and the streamed reply appear live to everyone watching. Conversations support branching (edit and resend), rename/archive, keyword and semantic history search, and promoting a personal chat into a shared project. - **Budgets scoped to projects, not just people.** Built for people who work across several engagements at once. Pre-flight enforcement blocks a send before it would exceed the project budget, with an 80% warning state; personal (per-user) and org-wide caps stack above as backstops. A blocked user can request an increase in one click, routed to an admin to approve or reject. - **Cost that reconciles to the customer's own books.** Each project carries the customer's own project/matter/engagement code, and that code carries through the token ledger into the analytics CSV export — so AI cost joins to existing client coding with no manual reconciliation step. - **Two ways to access models.** An organization can bring its own provider API keys (stored in Azure Key Vault, never exposed), use models ParleHub hosts with no key or provider account at all, or run both modes at once. Governance is identical either way: the same project budgets, personal and org caps, 80% warning, pre-flight hard block, token ledger, project code, and CSV export. - **Enterprise SSO and enterprise-native storage.** Microsoft Entra ID and Google Workspace OIDC sign-in, with domain-based auto-provisioning and optional SSO-only enforcement. Project files can live in the organization's own SharePoint site or Google Shared Drive so they never leave the customer's tenant; managed storage (Azure Blob) is available on every tier. - **AI that works the actual files.** The assistant makes real tool calls against project files: read, grep/glob/semantic search, exact find-and-replace and multi-edit, diffs, outlining, structured-data queries (JSON/YAML), spreadsheet work (Excel/CSV filter and aggregate, no code execution), and document conversion (PDF/Word/PowerPoint to markdown). Inline preview covers markdown, syntax-highlighted code, PDF, images/SVG, video, and sandboxed HTML. - **Ask the workspace about itself (Org Insight).** The assistant answers questions about live workspace state: a project's budget, spend and headroom, which models the organization may use, which external tools and file storage a project is connected to, and organization-wide spend. Figures are looked up fresh at the moment of the question rather than read from an index, so an answer is never a stale snapshot. In a shared project thread the assistant returns only what every member could already see for themselves, so organization-wide spend is answered in a personal conversation and withheld from shared ones; role entitlement still applies either way. The same rule applies to the prepaid credit behind hosted models: a project conversation is told whether it is healthy or exhausted, never the amount, because the amount is administrator-only on the screens that show it. Models that cannot call tools are marked at selection, with what stops working. - **Answers from the firm's own documents, with citations (Org Knowledge).** An organization-wide document corpus the assistant quotes with the document, version and section it came from, plus a link back. The primary path binds the corpus to a library the organization already maintains, a SharePoint site or Google Shared Drive reached through its existing storage connection, so the source of truth never moves and editing a document at source is what updates the answers; the corpus is reconciled against the source on a schedule, and administrators are alerted if a bound source becomes unreachable. Managed upload is available as a fallback for organizations with no library to bind. Retiring a document removes its passages from the index rather than hiding them, so a withdrawn policy cannot be quoted and an older citation to it resolves to nothing. The corpus is organization-wide, with a per-project opt-out. - **Governance and audit.** Four role levels (SystemAdmin, OrgAdmin, ProjectAdmin, ProjectMember), member invites, project visibility controls, and a sensitive-action audit log enforced append-only at the core platform level. Usage and spend analytics filter and group by project, person, provider, or model, with CSV export. ## Model providers Bring-your-own-key providers: Anthropic, OpenAI, Google Gemini (via the Gemini Developer API / AI Studio key), Google Vertex AI, and Azure OpenAI. Model spend on these lands on the customer's own provider contract at the customer's own rate. Vertex AI is connected with the organization's own Google Cloud credentials, so inference runs in its Cloud project, under its GCP contract, against its IAM, pinned to a region it chooses; which models a region serves varies, so the region is part of the setup rather than an afterthought. Hosted models run on Azure AI Foundry under ParleHub's managed identity, priced per token and drawn from a credit balance the organization tops up. The hosted catalog is deliberately curated rather than exhaustive, and it reaches beyond the three large model vendors to open-weight models a customer would otherwise have no contract with. Models can be switched per conversation. ## Who it is for Consultancies, law firms, accounting and advisory practices, agencies, and other organizations that bill by client or engagement and need AI usage centralized, cost-attributed, budget-capped, and auditable — rather than spread across individual accounts under terms nobody reviewed. ## Plans - **Free** — $0. 3 seats, 3 projects, model spend capped at $10/month, 2 GB managed storage, 10 MB max file size. - **Team** — $10/month flat. 10 seats, 10 projects, uncapped model spend, 10 GB managed storage, 25 MB max file size, Enterprise SSO. - **Enterprise** — $5 per seat per month, 10-seat minimum. Unlimited seats and projects, uncapped model spend, private (restricted) projects, SharePoint or Google Drive storage or 1 GB/seat managed, Enterprise SSO. Subscription pricing is separate from model spend. Model usage is billed either on the organization's own provider contract (bring your own key) or per token against a credit balance (models ParleHub hosts). The monthly model-spend cap is mode-neutral: it limits model spend whether tokens run on the customer's own keys or on models ParleHub hosts. Paid tiers include a 7-day trial. Billing is Stripe-backed. ## Pages - [Home](https://parlehub.com/): Positioning and core features — shared projects, project budgets, cost allocation by client code, model access, and governance. - [Security & Compliance](https://parlehub.com/security/): SSO, enterprise-native file storage, role-based access, the append-only audit trail, and platform hardening details. - [Pricing](https://parlehub.com/pricing/): Free, Team, and Enterprise plan comparison, plus FAQs on keys, hosted models, and spend caps. - [Blog](https://parlehub.com/blog/): Product notes and writing on AI governance, cost attribution, collaboration, and data residency. - [Blog archive](https://parlehub.com/blog/archive/): Chronological index of every post. - [Compare](https://parlehub.com/compare/): Hub for head-to-head comparisons against the model vendors' own team products. - [ParleHub vs Claude Team & Enterprise](https://parlehub.com/compare/parlehub-vs-claude/): Comparison on budget scope, cost allocation, shared threads, and model access. - [ParleHub vs ChatGPT Business & Enterprise](https://parlehub.com/compare/parlehub-vs-chatgpt/): Comparison on credit limits, shared projects, live collaboration, and model access. - [ParleHub vs Google Gemini & Vertex AI](https://parlehub.com/compare/parlehub-vs-gemini/): Comparison on Cloud Billing budgets, sharing model, and how ParleHub connects to Gemini. - [Terms of Service](https://parlehub.com/terms/): The agreement between ParleHub LLC and organizations using the service. - [Privacy Policy](https://parlehub.com/privacy/): What data is collected, how it is used, who it is shared with, and privacy rights. ## Blog posts - [Corrected at the Source: Why AI Should Point at Your Document Library, Not Copy It](https://parlehub.com/blog/2026/09/08/corrected-at-the-source/): Uploading policies into an AI tool creates a second copy that goes stale silently. Point at the library you already maintain, cite what you quote, and make retirement remove rather than hide. - [Looked Up, Not Remembered: Letting AI Answer Questions About Your Own Organization](https://parlehub.com/blog/2026/09/01/looked-up-not-remembered/): Governance dashboards hold the right numbers on a screen nobody opens mid-task. Live state has to be looked up rather than indexed, and a shared thread can only return what every member can see. - [The Budget Request Nobody Sends: What Happens When AI Spend Hits Its Cap](https://parlehub.com/blog/2026/08/25/the-budget-request-nobody-sends/): Hard limits on AI spend fail when there is no path through them; a cap needs a request-and- approve loop or people route around governance. - [Append-Only or It Didn't Happen: What Compliance Actually Asks of Your AI Log](https://parlehub.com/blog/2026/08/14/append-only-or-it-didnt-happen/): "We log everything" is not an audit trail. An audit trail is a record the application that wrote it cannot go back and change. - [A Project Name Is Not a Client Code: Why AI Cost Allocation Stalls Halfway](https://parlehub.com/blog/2026/08/12/a-project-name-is-not-a-client-code/): Most firms that set out to allocate AI spend to clients get the numbers right and the labels wrong, then reconcile by hand every month until they stop. - [Bring Your Own Key: Why You Shouldn't Rent Your AI Relationship](https://parlehub.com/blog/2026/08/09/bring-your-own-key-dont-rent-your-ai-relationship/): Renting model access is the right way to evaluate a model and the wrong way to run production work. Own the provider relationship; rent only the workspace. - [Cost Per Project Is a Product Decision, Not an Accounting One](https://parlehub.com/blog/2026/08/06/cost-per-project-is-a-product-decision/): Knowing what AI costs per project only matters if it changes what you build. - [Your AI Vendor Is a Data Residency Decision You Haven't Made Yet](https://parlehub.com/blog/2026/08/01/your-ai-vendor-is-a-data-residency-decision/): Every file pasted into a chat window is an unreviewed data-governance decision; the fix is keeping files in your own tenant. - [Why AI is Stuck in a Silo (and How Collaboration Fixes It)](https://parlehub.com/blog/2026/07/25/collaboration-in-ai/): Why isolated AI tools duplicate work, and what multiplayer AI with shared storage changes. - [Beyond the LLM Bill Shock: Why Enterprise AI Needs Project-Level Token Tracking and Budgeting](https://parlehub.com/blog/2026/07/24/why-enterprise-ai-needs-project-level-tracking/): Why traditional software cost governance fails for LLM spend, and what project-level tracking fixes. ## Contact - Sign up or sign in: https://app.parlehub.com - Legal: legal@parlehub.com - Privacy: privacy@parlehub.com ## Notes for retrieval - Every canonical URL on this site ends in a trailing slash; `.html` URLs redirect. - The canonical host is the apex domain `parlehub.com` (no `www.`). - A machine-readable page index is at https://parlehub.com/sitemap.xml. - ParleHub is not affiliated with or endorsed by Anthropic, OpenAI, Google, or Microsoft. Product names referenced on the comparison pages are trademarks of their respective owners.