QVeris
Run a task

QVeris Product Updates — Frequently Asked Questions

QVeris updates explain what changed, why it matters, and where to try it. The latest milestones make DeepSeek V4 Pro the Playground default, add guided setup for Cherry Studio and other desktop Agents, add full capability detail pages, let compatible OAuth clients request several resources in one consent flow, and smooth out plan upgrades.

What is the latest major QVeris update?
New Playground sessions now start with DeepSeek V4 Pro, whether the standard catalog or Model Gateway is in use. You can still choose another available model before running a task.
How do I connect Cherry Studio or another desktop Agent?
Open Plugins, choose your Agent, and copy the provided QVeris setup. Where remote setup is available, Hosted MCP is the recommended path; local stdio is available as a fallback.
What changed for API and SDK users?
Python, TypeScript, and REST references now have focused pages, search, outlines, copy actions, stable anchors, Markdown twins, complete operation coverage, and stable OpenAPI and llms indexes. The public Probe API also validates parameters and returns quotes without consuming credits.
What changed for CLI and agent authentication?
Terminal and headless workflows can use OAuth Device Flow without copying an API key. Persisted sessions, refresh and revocation, dynamic callback ports, and model-scoped Agent delegation keep client access explicit and bounded.
How are model and provider failures handled now?
Selectable models carry release evidence and tool-compatibility checks, while automatic provider routing avoids stale endpoint pins. Tool Finder also connects results to Inspect and Call with pre-call pricing and clearer non-billable rejection behavior.
How should I follow QVeris updates?
Use this What's New page as a lightweight changelog. Each item explains the user-facing change, why it helps, and where to try it.
PRODUCT UPDATES

What's New

A quick look at what changed, why it helps, and where you can try it next.

Reviewed through August 13, 2026 · 669 merged changes distilled into user-facing milestones

August 13, 2026
Feature

DeepSeek V4 Pro is now the default in Playground

New Playground sessions now start with DeepSeek V4 Pro, whether you use the standard catalog or Model Gateway. You can still choose another available model before you run a task.

Try DeepSeek V4 Pro →
August 12, 2026
Improvement

Plan upgrades now take you straight to checkout

Plan upgrades from Pricing or your account now use the payment link as soon as it is ready. This removes the extra wait that could leave an immediate upgrade without a way to finish payment, while scheduled changes remain manageable from your account.

View plans →
August 10, 2026
Feature

One OAuth approval can now cover your QVeris resources

Compatible OAuth clients can request account, tool, and model access together. QVeris returns separate resource-scoped tokens after one consent flow, so each token stays limited to the API it is meant to use.

Read the OAuth docs →
August 10, 2026
Feature

Cherry Studio and desktop Agents now have guided setup

The Plugins page now gives Cherry Studio, ChatGPT (Codex), Claude, Cursor, GitHub Copilot, Zed, and other supported desktop agents a copy-ready QVeris setup path. Where a remote connection is available, Hosted MCP is recommended first; local stdio remains a clear fallback.

Choose your Agent →
August 8, 2026
Feature

Explore every capability in a full detail view

Capability Map entries now open into shareable detail pages with task context, canonical inputs and outputs, provider paths, and an inline way to try a compatible tool. Deep links also work for catalog capabilities outside the curated scenario map.

Explore capabilities →
July 30, 2026
Feature

The new Application Center brings focused workflows together

Application Center is the new home for QVeris experiences built around complete workflows. The first collection includes Earnings Copilot for company and filing research and Options Assistant for structured market and strategy analysis. Rollout-gated applications join the same directory as they become available.

Explore the Application Center →
July 30, 2026
Feature

Tools and providers now share one discovery experience

Tool search and the provider directory now live at one Tool Discovery entry. You can search by need, browse callable tools, or compare providers without switching between disconnected landing pages. Existing search and provider links keep working while carrying their query and selected view into the unified directory.

Open Tool Discovery →
July 29, 2026
Feature

Turn a homepage task into a live result

The homepage now carries your task privately into Playground, preserves it through sign-in, checks model availability and credits, and can start the run automatically. Tool Finder, Provider Hub, Capability Map, and agent setup are also reachable within the first two viewports, while automatic provider routing keeps compatible models from being pinned to stale endpoints.

Run a task →
July 25, 2026
Release

API references now work for people and agents

Python, TypeScript, and REST references are now split into focused pages with search, page outlines, copy actions, stable anchors, and mobile-safe layouts. Every public Docs page also has a Markdown twin, all 20 REST operations have reference pages, and stable OpenAPI plus llms indexes keep machine-readable content aligned with the human docs.

Explore the new Docs →
July 23, 2026
Improvement

Trace usage back to the API key that made it

Usage history can now be filtered and attributed by API key, with billing redirects returning to the right account view. Search and Inspect responses were also separated into clearer contracts, and Inspect success now consistently returns HTTP 200, making audits and client handling less ambiguous.

Review usage →
July 21, 2026
Feature

Validate a tool call before spending credits

The public Probe API can validate parameters, coverage, and a price quote without invoking a provider or consuming credits. Typed schemas, rate limits, delegation checks, complete gateway errors, and the formal REST reference make the same preflight flow available to SDKs, CLIs, and agents.

Read the Probe API reference →
July 18, 2026
Feature

Sign in from a terminal without copying an API key

CLI and headless workflows can now use OAuth Device Flow, persist and revoke sessions, refresh tokens, and support dynamic local callback ports. Model-scoped Agent delegation and unified credential context keep each client limited to the access it was granted, while API-key authentication remains compatible.

Set up CLI authentication →
July 17, 2026
Release

Playground models are verified before release

The model catalog now carries release evidence, provider diagnostics, availability checks, and tool-compatibility gates before a model becomes selectable. Kimi K3 joined the validated catalog, and incompatible tool modes or unhealthy provider endpoints are filtered or routed around instead of failing after submission.

Try a verified model →
July 15, 2026
Feature

Official Hosted MCP is available

You can connect supported agents to QVeris through the official Hosted MCP service without running a local server. The rollout adds isolated endpoints, environment-scoped credentials, protocol-aware errors, readiness checks, security headers, exposure gates, and an auditable image supply chain.

Connect with Hosted MCP →
July 14, 2026
Improvement

Go from a search result to Inspect and Call

Tool Finder results now connect directly to Inspect and Call with responsive actions and the active language preserved. Playground separates free Discover and Inspect from paid Call, shows a compact pre-call estimate, blocks provider business rejections from being billed, and only offers models that support the selected tool flow.

Find and inspect a tool →
July 8, 2026
Feature

See which invitations turn into recharges

The invitation area now shows paid invitees, recharge totals, successful payments, granted credits, and drill-down payment records. Date filters use consistent boundaries and timezone handling so campaign and reward results are easier to reconcile.

Open invitation insights →
July 6, 2026
Improvement

Discovery results explain why they match

Discover responses and Tool Finder now surface categories, capabilities, and why each result was recommended. Instead of judging a match from a name alone, people and agents can see the routing rationale before inspecting or calling it.

Try explainable discovery →
June 29, 2026
Feature

Send product feedback without leaving the site

A shared feedback entry now lets you report an issue or suggestion from the product, attach a screenshot, and keep the report stored even when image upload is unavailable. The flow uses the same dark, localized experience across public pages.

Open Help Center →
June 25, 2026
Feature

Skill Hub grows into scenario-based workflows

Skill Hub now organizes 50 QVeris skills around real workflows instead of a flat catalog, including finance and supply-chain packs. Unpublished items stay out of the public catalog so every visible workflow is ready to use.

Browse Skill Hub →
June 25, 2026
Feature

Guides turn goals into practical workflows

A dedicated Guides index adds a direct path from a goal to a practical setup or implementation guide, with repaired guide routes and localized discovery.

Browse Guides →
June 22, 2026
Improvement

Homepage and Docs load more reliably

Homepage capability summaries now use resilient caching and static rendering, while route animation automatically scales back on lower-end devices. Docs gained static locale-aware rendering and fixes for first-load blank screens and language flashes, improving both speed and visual stability.

Explore QVeris →
May 28, 2026
Release

Browse providers by capability and quality

Provider Hub now supports category-level tool browsing, real provider identities, localized details, faster alias redirects, and comparable quality signals for similar providers. Capability Map moved to the complete v2 coverage contract, while the shared workbench keeps Inspect and execution context connected.

Browse Provider Hub →
May 22, 2026
Feature

Tool Finder and Provider Hub are live

You can now turn a plain-language need into the right QVeris tool, then inspect the provider behind it from one focused hub. It is a shorter path from discovery to provider context and the next call.

Try Tool Finder →
May 17, 2026
Release

Card recharge and auto-recharge are safer to resume

QVeris now carries the auto-recharge intent through Stripe more explicitly, keeps checkout card-only, handles canceled or returned sessions more cleanly, and resolves Stripe customer records before creating a new one.

Manage billing →
May 17, 2026
Improvement

Capability Map is easier to browse before sign-in

You can now explore the capability map and provider drawer before signing in. Authentication is still required when you try a live call, but comparing available finance capabilities no longer starts with a blocked page.

Open Capability Map →
May 17, 2026
Feature

Python SDK docs now use the right site examples

The Python SDK guide now has checked-in English and Chinese versions for each public docs experience. Examples use the right API base URL and display links for the current site, while install commands stay consistent.

Read Python SDK docs →
May 14, 2026
Release

Public API rate limits now fail clearer and earlier

The public REST API now applies rate limits before authentication on protected execute paths and returns rate-limit headers more consistently. Redis-backed limits also fail open where appropriate, so temporary limiter issues do not turn into broad API outages.

Review REST API docs →
May 7, 2026
Feature

Morningstar now has a dedicated provider qualification path

Teams evaluating Morningstar data now have a dedicated provider page and lead intake flow. The page highlights supported use cases, official branding, and the qualification path before teams start deeper integration work.

View Morningstar provider →
May 5, 2026
Improvement

Account history loads faster with cleaner controls

Usage and Credits Ledger pages load noticeably faster, especially when you're scrolling through longer histories. Both pages now share the same 1-day / 7-day / 30-day shortcuts and pagination, so switching between them feels consistent. Exports still grab the full filtered range you see on screen.

Review your usage →
May 4, 2026
Feature

Audit usage straight from your CLI or agent

Agents and the CLI can now check usage history and credit balances with the same API key they already use to call tools — no separate sign-in step. Ask short questions like "was that last call charged?" or "how much did I spend today?" and get a compact answer back instead of paging through everything.

Open Usage history →
April 29, 2026
Release

Execution outcomes now explain every charge

QVeris now records a standardized execution outcome for each tool call, separating transport success, provider success, result validity, and billable success. Usage & Execution History shows whether a request completed, failed, returned no usable result, or was not charged, while Credits Ledger stays focused on real balance changes with clearer quantity × unit price context. The goal is simple: every user should be able to tell what happened, why credits did or did not move, and where to audit it later.

Open Usage history →
April 25, 2026
Release

Every charge is now easier to understand

Your Usage page now puts each search, tool call, capability query, and model run in one place, with success status and charge status shown side by side. You can see the billing summary, trace it back to the ledger, and export the data, so it is much easier to confirm that failed requests were not charged.

Review your usage →
April 17, 2026
Improvement

Docs links now land where you expect

Links inside the docs now open the readable web page instead of dropping you into a raw markdown file. Setup guides, API key links, and related-doc blocks are easier to follow in both English and Chinese, while agent-friendly machine-readable routes still work.

Browse the docs →
April 17, 2026
Improvement

Choose the right setup path without guessing

The Plugins page now makes the recommended setup path clearer for each agent platform. Local agents point to the CLI flow, hosted builders point to MCP Server + skill, and the setup cards match the compatibility table, so installation takes fewer detours.

Open the Plugins page →
April 15, 2026
Improvement

Live Demo runs are easier to read after each call

Tool traces in the Live Demo now show the exact tool ID, use the same Call language as the rest of QVeris, and keep usage numbers tidy. After a run, you can quickly see what was called, which provider handled it, and roughly what it cost.

Open Live Demo →
April 15, 2026
Improvement

Plugin cards are easier to recognize

The main platform cards now use clearer dark-mode-ready brand assets. Hermes Agent, OpenClaw, Claude, OpenCode, Cursor, and other integrations are easier to spot, so you can get to the right setup guide faster.

Open the Plugins page →
April 13, 2026
Feature

Hermes Agent can call QVeris after one setup

If you use Hermes Agent, install the QVeris CLI once and you can call QVeris from its shell or terminal sandbox. Your credentials stay saved across sessions, so you do not need to repeat the same setup every time.

See the Hermes Agent install guide →
April 13, 2026
Release

Discover and call now mean the same thing everywhere

The Live Demo now uses the same wording your agents use: discover to find a capability, call to run it. That makes it easier to compare what happened in Live Demo, CLI, and MCP without translating between old and new tool names.

Open Live Demo →
April 10, 2026
Feature

Compare providers before you spend a call

Capability details now show success rate, latency, estimated cost, and call count for each provider. You can compare a few options first, then use Try-it to check whether the provider works for your real task.

Explore capabilities →
April 9, 2026
Improvement

Product posts are easier to share across teams

Blog posts now read cleanly in Chinese and English, with localized dates and better sharing metadata. It is easier to send the same guide to teammates without adding extra translation notes.

Read the blog →
April 8, 2026
Feature

Check finance coverage before you start integrating

Capability Explorer lets you browse and live-test finance capabilities across trading, market data, risk, research, alternative signals, and crypto. You can see what is covered and which providers fit before you spend engineering time.

Open the Explorer →
April 8, 2026
Improvement

The site is easier to navigate

The top navigation and footer are now grouped around the pages people usually look for: product value, developer setup, and ecosystem integrations. Whether you are new or returning, the right page should take fewer clicks to reach.

Browse the site →
April 7, 2026
Feature

Guides now read like web pages, not source files

Architecture notes, integration guides, and product updates now live in a readable blog experience with a table of contents, code highlighting, and responsive images. It is easier to learn the idea and then put it into practice.

Read the blog →
April 7, 2026
Release

Agents can call tools from the terminal with less context

The CLI lets agents call QVeris as a subprocess instead of carrying large tool schemas in context, which can cut prompt-token overhead by up to 80% compared with MCP. For local work, you also get REPL, JSON output, dry-run checks, and codegen.

Try the CLI →
April 7, 2026
Release

MCP tool names are simpler, and old setups still work

MCP now uses discover, inspect, and call as the main tool names, which better match how agents think about the workflow. Older tool names still work, and API key region handling remains automatic, so you can upgrade without breaking existing setups.

View MCP docs →
April 2, 2026
Feature

Pick tools with better signals, not guesswork

Search results and tool detail pages now show success rate, latency, and call count. When several providers can answer the same request, you and your agent have clearer signals before choosing one.

Try in Live Demo →
March 17, 2026
Improvement

It is harder to create duplicate accounts by mistake

Gmail dot-aliases and plus-tags now resolve to the same underlying inbox during registration. That helps avoid split accounts, missing credits, and confusing login states caused by slightly different versions of the same email.

Learn about security →
What's New — QVeris