MindsDB

MindsHub is a platform for team-oriented AI model access and cost management, with the core goal of centralizing the access, use, and spending of AI models into one place for management. According to the official introduction, it lets users use models in their own applications and agents through MindsHub Inference, or directly use models in MindsHub Agents to complete work, and uniformly manage access permissions, billing, and budgets through MindsHub Console. The product now uses the name MindsHub, and the page information is organized around its console, inference service, and agent capabilities.

Purpose and positioning The official description states that MindsHub solves the cost problem caused by out-of-control AI usage: when an AI experiment runs for a long time and more and more people, agents, and applications call models, the bill keeps growing, and teams often find it difficult to determine which work is valuable, which is wasting money, and who is responsible. Directly revoking a shared API Key may also interrupt services that others depend on, so teams need to first understand what is running and set appropriate access and spending rules. By centralizing permissions, model access, usage, and budgets in the console, MindsHub helps teams answer four questions: who can do what, which models can be used, where the money is spent, and when to intervene.

Main capabilities

1. Inference access: Access the model catalog through a unified MindsHub API Key, supporting integration with third-party agents or proprietary applications via the OpenAI Chat Completions or Anthropic Messages interfaces.
2. Access and permission management: Roles with the required permissions can be created, such as creating API Keys or viewing billing, without granting full administrator permissions; role and permission division at the organization, team, and workspace levels is supported, and workspace access is managed separately from organization roles.
3. Model access control: Access can be managed by provider or individual model, with rules set for teams or the entire organization; in the model access interface, cells can be blocked, allowed, or cleared, and a block rule in any container takes precedence, meaning that a member who belongs to two containers at the same time is blocked as long as they are blocked by either container; before any rules are added, everyone can use all models.
4. Usage and spending visibility: Daily spending by model, team, member, or API Key can be viewed, with drill-down into token usage and cost details; the dashboard shows the spending proportions of each model, member, team, and API Key.
5. Budgets and alerts: Budgets can be set for teams, and models can be selectedand budget cycles and track quota usage; the Alert only option will alert when the quota is reached but keep requests running, and by default paid requests are paused when the budget reaches 100%.
6. Audit logs: record who in the organization changed what, when, and with which credentials, showing the most recent 90 days, with a date filter available to narrow the range.
7. Request and session tracking: you can view every inference request sent by the account and the session it belongs to, and it provides metrics such as request volume, error rate, p50/p95 latency, and streaming latency; sessions come from agent traffic, and direct API calls do not carry sessions.

Agents and Models MindsHub Agents includes two modes: Cowork and Code. Cowork is oriented toward research work, and is officially described as "an agent workspace for knowledge work," capable of research, analysis, and creation; Code is a Codex-based coding agent used to build features, fix bugs, and review code. On the model side, MindsHub Air is a model for non-coding everyday knowledge work, and its underlying model may change as better options appear. Jev is TypeSafe's decision model, returning options, scores, and probabilities rather than chat replies, and can be called through the Decisions API under the name jev together with a MindsHub API Key.

Billing Methods Officially, there are two methods: a permanent free tier and pay-as-you-go. The free tier requires no credit card and includes the MindsHub Air model for non-coding knowledge work, classification/routing/scoring through Jev, and a daily quota within fair use limits; the official guidance is that batch automation or high-traffic workloads should use pay-as-you-go. Pay-as-you-go is billed at provider rates, uses prepaid balance, has no expiration or minimum spend, includes the full model catalog, cache savings (when repeatedly sending content already cached by the provider, billing is at the lower cached input rate, provided the provider offers that rate), no plan limits, automatic top-up with a monthly cap, and usage and spend tracking.

Applicable Scenarios Suitable for teams that need unified management of model calls and spending across multiple people, multiple agents, and multiple applications, for troubleshooting waste, reviewing access permissions, deciding how to adjust spending direction, and connecting their own applications or third-party agents with a unified key.

Other Notes The official page mentions the use of cookies and similar technologies, with necessary cookies always enabled, and analytics (PostHog server-side) and marketing measurement cookies used after consent, and states that user information will not be sold or used to build advertising profiles.For specific prices and model rates, please refer to the official pricing page.