The Sovereign Governance: Deterministic Rules & Compliance Protocol that secures your AI Agents.
Intercepts intent, enforces policy, and generates immutable audit trails.
11 integrated capabilities — from Intent (CHI) to Consensus (Quorum) — plus three Core Vaccines for runtime defense.
Pre-flight regulatory and cost governor. Blocks policy-violating actions before any network call — offline, no round-trip required.
Forces a structured decision cycle before any irreversible write operation — no shortcuts.
High-performance Rust Enforcer (<5ms) running in Edge sandbox.
Mandatory intent introspection before execution. Detects semantic drift, PII, and prompt injection in real time.
Memory compliance and data state introspection to sanitize long-term memory streams.
SHA-256 forensic ledger with Bitcoin or L2 anchoring. Every decision is cryptographically verifiable.
Pre-built YAML policy packs for HIPAA, SOC2, LGPD, and financial compliance. Accelerates audit preparation.
Enforce multiple M-of-N signatures for high-risk decisions. Zero-Trust approval chains.
Automated authority syncing via Webhooks. Inherit corporate hierarchies into governance.
Intercept every tool call before execution. JSON/YAML Enforcement for AI agents.
Define policies in JSON/YAML instead of code. 7 built-in system policies block credential theft, path traversal, and exfiltration.
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.
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" });// Single-agent, proxy-compatible frameworks. Multi-agent (CrewAI, LangGraph) requires the @abs-core/sdk.
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.
Financial Accountability: Agents post collateral against policy compliance. Violations trigger automatic slashing — making risk quantifiable and insurable.
Enterprise 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 // Evaluate before executenpm install @abs-core/sdk # Production Governance