Research
AI Agent Execution PythonOpen source · Active

Lattice Capability Runtime

An open-source runtime for structured AI-agent execution through explicit capabilities, bounded authority and controlled interaction with enterprise systems.

Overview

Lattice is a capability runtime for outcome-based execution. It sits between an AI agent and the systems it acts on, so that sequencing, state, credentials, failure policy and audit live in reviewed code rather than inside the model's context window.

Instead of exposing an open-ended tool surface, Lattice exposes typed capabilities with explicit inputs, outputs, permission boundaries and execution tracing. The agent expresses what it wants to achieve; the runtime executes deterministically and returns a compact projection the model can reason over.

This keeps the agent in the role of interpreting intent and applying policy, while the runtime remains accountable for how work is actually performed against enterprise systems.

Architecture diagram
AI Agent
Expresses an objective and reasons over projections
Lattice Capability Runtime
Typed capabilitiesBounded authorityDurable stateExecution trace
Enterprise Systems
APIs, data stores and services with real credentials

The agent submits an objective and reasons over projections; Lattice governs execution — capabilities, authority, state and audit — before touching enterprise systems.

Core concepts
1

Typed capabilities

Operations are declared with explicit schemas, side-effect classification and required authority, replacing an unbounded tool surface with a reviewed set of actions.

2

Bounded authority

Each capability runs within a scoped permission envelope, so an agent cannot exceed the authority granted for the task at hand.

3

Durable state

Intermediate state is held by the runtime rather than reconstructed from context, keeping long or resumable tasks consistent and inspectable.

4

Execution tracing

Every capability invocation is recorded with inputs, outputs and outcomes, providing an audit and replay trail for what the agent caused.

5

Projections

The runtime returns a structured, minimal view of results for the model to reason over, limiting context bloat and tool-space interference.

Execution flow
1

Intent

The agent submits an objective and the capabilities it intends to use.

2

Authorization

The runtime checks requested capabilities against the task's bounded authority.

3

Execution

Capabilities run deterministically inside the runtime, holding state and credentials.

4

Evidence

Inputs, outputs and outcomes are traced for audit and replay.

5

Projection

A structured result is returned to the agent to inform its next decision.

Security model
1

Reviewed execution path

Sequencing, credentials and failure handling live in code that can be reviewed, tested and versioned.

2

Least authority

Capabilities carry only the permissions they need, scoped per task rather than granted globally.

3

Side-effect classification

Capabilities declare whether they read, write or cause external effects, enabling policy before execution.

4

Containment

A compromised or misaligned model cannot exceed the capability boundary or fabricate execution.

Source repository

csehammad/Lattice

The implementation is developed openly on GitHub — Python. The repository is the source of truth for code, examples and documentation.

Open on GitHub
Related whitepaper

Outcome Coordination for AI Agents: Governing Execution Across Trust Boundaries

Lattice is the reference implementation for execution governance within a single trust boundary in the Outcome Coordination architecture.

Read the paper