Python multi-agent runtime that ships Jev natively in an optional extras install: agents ask typed Choice, Score and Noul questions through a shared decision client, the Kernel can pick specialist subagents by Choice, and RAG passages with a prompt-injection Noul above 0.70 are withheld.
JarvisCore is a runtime where AI agents operate as a fleet of equal peers. Agents discover one another by capability, execute tasks from a shared ledger, and authenticate to every external service through a zero-trust broker, never with their own keys.
Repo stats from the GitHub API, cached Oct 2026.Homepage
About this project
The production runtime for multi-agent systems — a peer-to-peer mesh with no central orchestrator, zero-trust credentials, durable state that survives kill -9 , and a built-in catalog of typed integration atoms. Observability included, not upsold. Watch the JarvisCore Agents launch film.
pip install "jarviscore-framework[typesafe]" adds a shared async Jev client exposed as self.decisions
KERNEL_ROUTER_PROVIDER=typesafe makes Jev pick specialist subagents with a Choice question
RAG_DECISION_PROVIDER=typesafe classifies the FAISS shortlist into accepted, conflicting and excluded passages
A prompt-injection Noul above 0.70 withholds the RAG passage from the generating model
AutoAgents get an evaluate_decisions thinking tool; Jev complements, not replaces, the generative model
Watch out
Apache-2.0. The Jev features need TYPESAFE_API_KEY and Python, and the framework is much broader than Jev, which is one of several decision providers.
Questions about this entry
Short answers drawn from the linked project or page.
No. Jev is an optional TypeSafe extras install; the runtime works with other LLM providers, and the Jev-backed router, task-complexity and RAG decision providers are opt-in via environment variables.
Agents call self.decisions.evaluate with typed questions, such as a Choice of owning team with criteria per label; answers come back per question with probabilities. AutoAgents can also use the evaluate_decisions tool.
Install the TypeSafe extra with pip install "jarviscore-framework[typesafe]", which adds a shared async Jev client exposed as self.decisions, and supply a TYPESAFE_API_KEY.
Set KERNEL_ROUTER_PROVIDER=typesafe and the Kernel router uses a Jev Choice question to pick a specialist; RAG_DECISION_PROVIDER=typesafe does the same for classifying the FAISS shortlist.
A prompt-injection Noul scoring above 0.70 withholds that passage from the generating model, so injected instructions in retrieved text do not reach the generator.
It is Apache-2.0 and Python. The Jev features need TYPESAFE_API_KEY, and Jev is one of several decision providers rather than a requirement.
In Redis: workflow state, step outputs and traces live there, so a process can be killed with kill -9 and rerun on the same workflow id; completed steps recover from Redis instead of running again.
Alibaba's multi-agent platform, which implements native TypeSafe Jev classification through JevClassifierModel for binary, choice and score routing across agent pipelines. Python 3.11+, Apache-2.0, with a FastAPI agent service and web UI.
Autonomous agent platform with 187,504 stars for building and running agents from plain English or a visual builder. Its platform features first-class TypeSafe Jev decision blocks for typed routing, filtering, scoring and confidence-gated next-action dispatching.
Python agent framework whose TypeSafeModel class runs Jev as a decision model inside Pydantic AI agents: Pydantic output-type fields become typed Jev questions answered together in one request, each answer carrying its own confidence, and tools can be offered as Choice options.
Open-source TypeScript agentic operating system: a core runtime, app for web and mobile, CLI, cloud services and first-party plugins. The directory's curated summary reports it integrates TypeSafe System One decision services for sub-100ms intent classification, action dispatching and confidence-gated tool execution.
Open-source AI email assistant (Next.js, Postgres, 12k+ stars) with a decision-model layer that can run on openrouter:typesafe/jev-1.13: Jev-backed questions cover rule selection, cold-email detection, thread status, recurring patterns, unsubscribe pages, and sender categorization, each with evals.
