Qveris Agent Toolkit
Open-source toolkit for the QVeris capability routing network: CLI, MCP server, Python SDK, skills, and REST API docs for agents to discover, inspect, call, and audit real-world tools.
Last verified:
What is Qveris Agent Toolkit?
Qveris Agent Toolkit is a capability routing network for AI agents that enables them to discover, inspect, and call real-world, verified capabilities through one unified protocol. It provides access to 10,000+ capabilities across 15+ categories including quant trading, macro/fixed income, investment research, risk/compliance, crypto digital assets, and alternative signals like news and social media.
The toolkit operates on a three-step workflow: Discover (find capabilities using natural language queries), Inspect (check parameters, examples, latency, success rate, and estimated cost), and Call (execute in a sandboxed environment and receive structured machine-readable JSON output). Discover and Inspect are always free, while Call consumes credits based on the capability's billing rules.
Qveris is built for AI agent developers and teams building production-grade agents who need access to live data, external APIs, and third-party services without hardcoding provider lists. It supports 14+ agent platforms including Claude Code, Cursor, CLI, CPP, Python SDK, REST API, Hermes Agent, and OpenClaw. The platform includes Provider Hub for browsing provider coverage, quality signals, and audit trails for production calls.
Key features include dynamic tool discovery by capability description, smart failover with multiple provider routes, edge nodes with under 500ms p95 latency, 99.99% uptime SLA, sandboxed execution environment, full audit trails, RBAC access control, CLI for terminal access, MCP Server integration, and Python SDK for programmatic access.
Qveris Agent Toolkit pricing
Pricing model: Freemium
Qveris uses a credit-based pay-per-use model with no monthly subscriptions. Free tier includes 1,000 credits on signup plus 100 daily login credits that reset each day, with 10 requests per minute limit. Pro plan costs $19 for 10,000 credits with 100 req/min and $0.002/credit overage. Scale On-Demand plan starts at $1 minimum with bundle options: $100/$500/$1,000 with 5%/10%/15% bonus credits. Example costs: Get a live quote ~1 credit, OCR a page ~2 credits, Parse a PDF 3-10 credits, Analyze a financial report 5-15 credits, Generate an image 5-20 credits. Discover and Inspect are always free; only Call operations consume credits based on each capability's billing rule.
Qveris Agent Toolkit pros
- 10,000+ capabilities across 15+ categories
- Discover and Inspect are always free
- 1,000 credits free on signup
- 100 daily login credits that reset each day
- Supports 14+ agent platforms including Claude Code and Cursor
- Natural language tool discovery without hardcoded providers
- 99.99% uptime SLA for production reliability
- Under 500ms p95 latency with edge nodes
- Sandboxed execution environment for safety
- Full audit trails for debugging and governance
- Smart failover with multiple provider routes
- Provider Hub with quality signals and latency metrics
- Structured JSON output for downstream systems
- CLI for command-line tool discovery and execution
- MCP Server and Python SDK available
- No monthly subscription fees - pay per use
- RBAC access control for enterprise security
- Session-scoped configuration for tracing
Qveris Agent Toolkit cons
- Only 10 requests per minute on free tier
- Credits required for Call operations
- No traditional monthly subscription plans
- 1 credit minimum for top-up purchases
- Limited to finance-focused capability domains primarily
- Provider availability varies by capability
- Cost estimation only pre-settlement, final billing from ledger
- Requires API key setup and environment variables
- Learning curve for Discover-Inspect-Call workflow
- Session ID not a cache contract with no cache hit guarantee
Frequently asked questions about Qveris Agent Toolkit
What is Qveris and what does it do?
Qveris is a capability routing network for AI agents that lets your agent discover, inspect, and call real-world, verified capabilities through one unified protocol. It provides access to 10,000+ capabilities across categories like finance, weather, search, and data retrieval. Agents can search for tools using natural language descriptions, inspect candidates for latency and cost, then execute them with structured parameters to get machine-readable JSON output.
Is Discover and Inspect really free?
Yes, Discover and Inspect are always free. Discover returns a search_id plus ranked capability candidates without consuming credits. Inspect provides full parameter details, examples, quality metrics, and billing rules before execution, also free. You only pay credits when you Call (execute) a capability and get real value from the results.
How do I get started with Qveris?
You can start in 30 seconds: sign up to get an auto-provisioned API key with 1,000 free credits, then pick your integration path. Options include installing the CLI with curl, setting up MCP Server for your agent client, using the Python SDK with pip install qveris, or calling the REST API directly at https://qveris.ai/api/v1. Your agent then self-configures with Qveris to discover and execute capabilities.
What agent platforms does Qveris support?
Qveris supports 14+ agent platforms including Claude Code, Cursor, CLI, CPP, Python SDK, REST API, Hermes Agent, and OpenClaw. Each platform has specific setup instructions, and the MCP Server provides a unified way to add Qveris capability to any MCP-compatible client.
How does billing and credits work?
Discover and Inspect are free. Call operations consume credits based on each capability's billing_rule. You start with 1,000 signup credits and get 100 daily login credits that reset each day. Example costs range from ~1 credit for a live quote to 5-20 credits for image generation. The Call response includes pre-settlement cost signals, but the usage audit and credits ledger are the source of truth for final settlement.
What happens if a provider is unavailable?
Qveris provides smart failover with multiple provider routes. One capability can map to multiple providers, so when one provider is unavailable or priced poorly for the task, Qveris exposes alternative supply. Agents start from the capability goal instead of being dependent on a single API, ensuring production reliability with 99.99% uptime SLA.
How do I inspect a capability before calling it?
Use the Inspect step to check parameters, examples, latency, success rate, and estimated cost before execution. You can call inspect with a tool_id and search_id to get full details: latency (~180ms), success rate (99.8%), and cost (3 credits). This helps your agent pick the best candidate using parameters, success_rate, latency, region, and billing_rule before committing to a Call.
What categories of capabilities are available?
Qveris offers 15+ capability categories including quant trading (index constituents, factor data), macro/fixed income (yield curves, inflation prints), investment research (earnings, analyst consensus, valuation), risk/compliance (KYC, sanctions screening), crypto/digital assets (on-chain activity, DeFi protocols, stablecoin flows), and alternative signals (news, events, social media). The capability map shows coverage before setup.
How do I audit and debug agent calls?
Qveris provides full audit trails for production calls. Every call includes execution_id, search_id, and session_id for tracing. You can query usage audit history at /auth/usage/history/v2 with your execution_id, or check the credits ledger at /auth/credits/ledger. Call results, provider choice, cost, and execution context become queryable records for debugging, review, and governance.
What is the recommended agent loop for Qveris?
The recommended workflow is: 1) Discover with a capability-level query (not final parameters), 2) Inspect 1-3 candidates when schema or cost matters, 3) Pick the candidate using parameters, success_rate, latency, region, and billing_rule, 4) Call with structured parameters, 5) Use execution_id, search_id, and session_id when auditing or debugging. This loop works in agent loops with tool discovery → execution → feed results back to the model.