The Sovereign Governance: Deterministic Rules & Compliance Protocol that secures your AI Agents.
Intercepts intent, enforces policy, and generates immutable audit trails.
12 integrated capabilities — from Intent (CHI) to Pre-Auth (PAP) — plus four Core Vaccines for runtime defense.
Pre-flight regulatory and cost governor. Blocks policy-violating actions before any network call.
Forces a structured decision cycle before any irreversible write operation.
High-performance Rust Enforcer (<5ms) running in Edge sandbox.
Mandatory intent introspection before execution. Detects semantic drift and PII.
Maintains participant reputation scores and detects logic drift in real time.
SHA-256 decision chain with Bitcoin or L2 anchoring for immutable auditing.
Pre-built YAML packs for HIPAA, SOC2, and financial compliance readiness.
Enforces M-of-N human approvals for critical tier-1 transaction paths.
Okta & LDAP integration via Webhooks. Inherit corporate hierarchies into docs.
Secure gateway for Model Context Protocol tools and RCE protection.
Declarative JSON/YAML rule definition for behavioral governance.
Air-gapped verification utilizing PAAT tokens to ensure offline constraint adherence.
Proactive defense system protecting your agent's runtime.
Prevents key leakage by injecting credentials only at runtime. Your agent never sees the API key.
Health check and recovery monitoring for autonomous agents. Triggers critical alerts if the agent stops reporting.
Run policies in observation mode. Measure risk before enforcing blocks.
Air-gapped verification utilizing PAAT tokens to ensure offline constraint adherence.
Works with your existing stack.
Drop-in integration via @abs-core/adapter-langchain. Intercepts LLM/Tool start events.
Governance for autonomous Python agents via the abs-sdk. The Magic Proxy covers simple single-agent calls; multi-agent frameworks (CrewAI, LangGraph) require explicit SDK integration.
Zero-code adoption for Next.js apps via @abs-core/vercel-adapter. Seamless Protocol Translation.
const openai = new OpenAI({ baseURL: "https://api.abscore.app/v1" });Decoding the Immune System Acronyms.
Offline Cost Sentinel: Pre-flight regulatory and cost governor. Blocks policy-violating and cost-bearing actions before any network call — offline, no round-trip required.
Cognitive Host Interface: Intent introspection engine. Verifies reasoning before state changes.
AI Change Control Protocol: Forces a structured decision cycle before any irreversible write operation — no shortcuts.
Universal Kernel: High-security Rust sandbox that executes policy checks without agent environment escape. Evaluation in < 5ms.
Infrastructure Governor: Specialized vaccine for infrastructure-as-code. Blocks unauthorized cloud resource mutations before any API call.
Declarative Rules: High-level governance for infrastructure operations (files, shell, cloud APIs).
Financial Accountability: Agents post collateral against policy compliance. Violations trigger automatic slashing — making risk quantifiable and insurable.
Pre-Authorization Protocol: Air-gapped, fail-closed verification utilizing cryptographic PAAT tokens to ensure constraint adherence even in isolated network environments.
Sovereign governance scaled for everyone.
See the Magic Proxy blocking an agent trying to read .env or exfiltrate AWS credentials, while allowing secure log reading operations.
Intercept every tool call before execution. Policy-as-Code for AI agents.
Intercepts MCP tool calls with layered policy evaluation, entropy analysis, and immutable audit logging.
Define policies in JSON instead of code. 7 built-in system policies block credential theft, path traversal, and exfiltration.
HMAC-SHA256 tokens with tool permissions, risk budgets, and bond status. Low-latency verification at the edge.
POST /v1/mcp/tool-call npm install @abs-core/sdk