Open-source tools for AI agent decisions, memory, behavioral contracts and bounded execution. Qualixar is an independent research initiative by Varun Pratap Bhardwaj.

10 core products · 10 public arXiv preprints

Jev returns advisory decisions; its answer does not authorize an action. Bounded Loops' first keyless example uses a stub worker and real pytest checks. SuperLocalMemory's optional providers and connectors have separate network behavior. AgentAssert's small example checks a caller-supplied flag; it is not a security classifier.

# Bounded Loops: install a loop in your project and run its keyless example
python -m pip install bounded-loops
bl loops install bug-fix-red-green --dest ./loops
bl run loops/bug-fix-red-green --yes --run-id first-proof

# SuperLocalMemory: install, then select your operating mode explicitly
npm install -g superlocalmemory
slm setup

# AgentAssert: install the YAML and mathematical dependencies
python -m pip install 'agentassert-abc[yaml,math]'

For Jev, follow the host-specific installation guide. Check each repository's supported platforms and setup requirements before installation.

Read Qualixar's research directory alongside each product's source and examples. Public arXiv papers are preprints; an arXiv record does not establish peer review. Mathematical results and experiment figures apply to the assumptions, datasets and versions described in each paper.

• Bounded Loops — arXiv:2609.27871
• SuperLocalMemory 4.0 — arXiv:2608.08253v2
• Agent Behavioral Contracts — arXiv:2602.22302

Use the Qualixar community hub:

• Q&A: installation, configuration and usage questions.
• Ideas: describe a task and the workflow you want to improve.
• Polls: help prioritize tutorials and examples.
• Announcements: maintainer updates and release links.
• Show and tell: share an integration and what you observed.

Report reproducible bugs in the affected product's issue tracker. Follow the repository's published security-reporting instructions where available; do not post vulnerabilities, secrets, private prompts or customer data publicly. See the community guide.

If a product helps your work, starring its repository is one way to follow it. Using the tools and participating in discussions does not require a star.

Choose the capability your workflow needs, then follow the setup, examples or research in Resources. Product names link directly to the source repositories.

Use each repository’s setup guide for supported hosts and operating modes. Benchmark results apply to their published datasets and protocols; the SkillFortify benchmark linked above uses generated specimens.

These public arXiv records are preprints. Read each paper for its assumptions, version, experimental scope and limitations. Listing a paper beside a product does not establish that every current feature was evaluated in that paper.

Qualixar · SuperLocalMemory · AgentAssert · Author and research