Back to Apps

VAKRA
by ibm
Enterprise-grade benchmark for evaluating multi-hop, multi-source tool-calling in AI agents.
0 stars
Works in:claude
Exposes:ToolsResources
What it does
VAKRA (eValuating API and Knowledge Retrieval Agents) is a tool-grounded executable benchmark designed to evaluate AI agents' compositional reasoning in complex enterprise environments. It measures how well agents can chain decisions across multiple APIs and documents, reconciling mismatched schemas and following natural-language tool-use constraints.
Tools
VAKRA provides an executable environment with over 8,000 locally hosted APIs across 62 domains. These are categorized by capabilities:
- API Chaining: Nested and compositional API calls.
- Tool Selection: Large-scale selection over query-aligned endpoints.
- Multi-hop Reasoning: Dependent reasoning chains where earlier outputs parameterize later calls.
- Multi-turn Reasoning: Complex dialogues combining APIs and document retrieval with policy adherence.
Installation
Requires Docker and Python 3.11+. Run the following:
make download
make build
docker compose up -d
Supported hosts
- claude
Quick install
make download && make build && docker compose up -dInformation
- Pricing
- free
- Published
- 8/4/2026
- stars
- 0
Categories
Choose your AI client and follow the steps below.
Claude Desktop
Follow the repository guide: make download, make build, and docker compose up -d to start the benchmark environment.






