WAIL v1.3 / Execution Control & Governance

Control what you can’t control.
Prove what actually happened.

WAIL is an AI, Agent & MCP Execution Control & Governance Layer for production AI systems. It observes execution, detects degradation, supports controlled interventions, enforces execution governance, and preserves cryptographically signed evidence.

No gateway. No replacement agent framework. Designed for on-premises and private-cloud environments.

Explore WAIL Runtime Explore verification Evidence Dataset
01 / Platform capabilities

One layer across AI execution

Runtime control and governance across model calls, agent runs, and MCP tool operations.

RUNTIME

Observability & Detection

Capture execution signals, establish behavioral baselines, and identify runtime degradation, anomalies, and risk.

  • Latency and first-token behavior
  • Provider and model execution signals
  • Execution risk and anomaly detection
CONTROL

Controlled Runtime Actions

Support evidence-driven observe, retry, and reroute decisions within the applicable execution scope.

  • Policy-constrained intervention
  • Controlled provider reroute
  • Recovery outcome verification
AGENT & MCP

Connected Execution Graphs

Represent Agent Runs, model execution, and MCP tool operations as connected execution units.

  • Agent, Model, and MCP Tool Units
  • Execution relationships and chronology
  • Traceability across multi-step activity
GOVERNANCE

Continuous Governance Enforcement

Evaluate authorization, policy, and authority changes throughout active execution, not only at request start.

  • Mid-execution authority changes
  • Intervention in affected active operations
  • Blocking unauthorized subsequent dispatch
EVIDENCE

Signed Execution Evidence

Connect runtime observations, decisions, control actions, and outcomes through cryptographically signed records.

  • RSA-PSS-SHA256 evidence signatures
  • Chronological governance records
  • Ledger-linked verification workflows
DEPLOYMENT

Customer-Controlled Infrastructure

Integrate around existing AI applications without replacing their orchestration frameworks.

  • On-premises and private-cloud design
  • Provider-independent execution control
  • Customer-controlled runtime data
02 / Execution lifecycle

From observation to proof

WAIL distinguishes what the system observed, what it decided, what action was attempted, and what outcome was verified.

OBSERVE DETECT DECIDE ACT VERIFY PROVE
A recorded decision is not proof of enforcement. WAIL distinguishes a control decision, attempted intervention, local withholding, and independently verified remote termination. Evidence must accurately reflect the observed outcome.
03 / Verification & Evidence

Technical verification domains

WAIL v1.3 verification focuses on execution behavior, governance enforcement, controlled recovery, and evidence integrity. Each result is interpreted within its tested scope.

V-01 / GOVERNANCE

Mid-execution REVOKE

Authorization changes during active execution and enforcement across affected operations.

Verification domain
  • Active-operation intervention
  • Unauthorized dispatch prevention
  • Local versus remote cancellation distinction
V-02 / EXECUTION GRAPH

Agent → MCP → Model

Connected execution identities and signed records across agent, tool, and model activity.

Verification domain
  • Agent Run relationships
  • MCP tool execution evidence
  • Cross-unit chronological traceability
V-03 / RECOVERY

Controlled Reroute

Controlled runtime reroute and verification of the resulting execution behavior.

Verification domain
  • Provider reroute scenarios
  • Agent-scope control
  • Action and outcome separation
V-04 / EVIDENCE

Signed Evidence & Ledger

Cryptographic integrity and chronological linkage of execution and governance evidence.

Verification domain
  • Signature and chain verification
  • Checkpoint and trusted-head validation
  • Evidence commit failure handling
Evidence verification: WAIL v1.3 verification artifacts are organized in the official Evidence Dataset, including test reports, signed records, integrity manifests, and reproduction instructions as they are published. This static page does not execute tests.
Open WAIL v1.3 Evidence Dataset
04 / Developer access

Start with WAIL Runtime

Explore the public developer runtime and its integration documentation. 

pip install wail-runtime

from openai import OpenAI
import wail

client = wail.wrap(OpenAI())
PyPI GitHub Official website