Skip to content

warheart1984-ctrl/agentic-coding-agent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Nova: Constitutional Agentic Coding System

A governed, auditable, multi-agent coding platform with cryptographic provenance, lineage tracking, and constitutional governance.

Repository: warheart1984-ctrl/agentic-coding-agent
Mission: #002 — Governed Agentic Coding & Verification
Kernel: CRK-2 — Constitutional Runtime Kernel v2

License: MIT Mission Kernel


What It Does

Nova runs AI agents that write code under constitutional governance — every action produces cryptographic receipts, every decision has provenance, every state change is auditable. Think claude-code or opencode but with constitutional invariants, Merkle receipts, and full causal lineage built in.

Core guarantees:

  • Lawful execution — Agents cannot violate constitutional invariants
  • Cryptographic receipts — Every tool call, decision, and state transition is signed
  • Causal lineage — Full reconstruction of why any state exists
  • Multi-agent orchestration — Controlled coordination via Control Tower
  • Replay & verification — Observer bundle enables independent reproduction

Quick Start

# Prerequisites: Node 20+, pnpm, Docker
git clone https://github.com/warheart1984-ctrl/agentic-coding-agent
cd agentic-coding-agent

# Install & build
pnpm install
pnpm build

# Start the stack (API, Cockpit, Kernel)
docker compose up -d

# Run a governed agent
pnpm agent:run --plan ./examples/refactor-auth.yaml --workspace ./my-project

# Observe in Cockpit
open http://localhost:8080

Docker quickstart: See QUICKSTART-DOCKER.md
Full deployment: See DOCKER-DEPLOYMENT.md


Architecture

┌─────────────────┐     Intent + Receipts      ┌──────────────────────────┐
│   API Server    │ ─────────────────────────▶ │  Constitutional Kernel   │
│   (Ingress)     │   Governed Execution       │  (CRK-2: Invariants,     │
└─────────────────┘                            │   Policies, Receipts)    │
                                               └────────────┬─────────────┘
                                                            │ Governs
                                                            ▼
┌─────────────────┐                            ┌──────────────────────────┐
│   Cockpit UI    │ ◀─── Observes ─────────── │     Agent Runtime        │
│  (Observability)│   Receipts, Lineage,      │  (Agents, Tools, LSP,    │
└─────────────────┘   Diffs, Plans            │   Control Tower)         │
                                                └────────────┬─────────────┘
                                                             │ Evidence
                                                             ▼
                                                  ┌──────────────────────────┐
                                                  │    Persistence Layer     │
                                                  │  (WAL, Snapshots, CSR,   │
                                                  │   Observer Bundle)       │
                                                  └──────────────────────────┘
Layer Responsibility Key Components
API Server gRPC/REST ingress, auth, rate limiting backend/
Constitutional Kernel Invariants, policies, receipts, lineage crk2/, src/kernel/
Agent Runtime Multi-agent loops, tool exec, LSP agent/, src/runtime/
Control Tower Multi-agent orchestration, coordination control-tower/
Tool Subsystem File/Git/LSP/Exec/Web/Skills tools/, src/tools/
Persistence WAL, snapshots, causal reconstruction src/persistence/, .sxk-wal/
Cockpit Real-time plan/diff/receipt/lineage viz cockpit/

Constitution

Governance is defined in config/constitution.yaml:

invariants:
  - "no-secrets-in-code"
  - "all-changes-require-tests"
  - "no-force-push-to-main"

policies:
  - name: "code-review-required"
    condition: "diff.lines > 100"
    require: "approval-from: @maintainers"
  
  - name: "crypto-no-rollback"
    condition: "operation.touches: crypto"
    forbid: "state-rollback"

receipts:
  required: ["tool-invocation", "decision", "state-transition"]
  retention: "7y"
  anchoring: "sha256+timestamp"

Key concepts:

  • Invariants — Hard constraints the kernel enforces (never violated)
  • Policies — Conditional rules requiring approval or forbidding actions
  • Receipts — Cryptographic proof of every execution step
  • Lineage — Causal chain from intent → plan → execution → outcome

Key Commands

# Agent execution
pnpm agent:run --plan <plan.yaml> --workspace <path>
pnpm agent:run --interactive --workspace <path>

# Observability
pnpm cockpit:dev          # Start Cockpit UI (port 8080)
pnpm receipts:verify --session <id>
pnpm lineage:show --session <id> --format mermaid

# Governance
pnpm kernel:validate --constitution config/constitution.yaml
pnpm policy:test --plan <plan.yaml>

# Verification
pnpm observer:replay --bundle observer-bundle-*.zip
pnpm test:conformance       # CRK-2 conformance suite

Documentation

Guide Description
Architecture Deep dive on kernel, runtime, persistence
Constitution Guide Writing policies, invariants, receipts
Cockpit UI Observability dashboard
API Reference gRPC/REST schemas
Skills Development Building custom tools
CRK-2 Spec Constitutional Runtime Kernel v2
Mission #002 Governed agentic coding objectives

Verification & Evidence

This repository produces constitutional execution records:

  • Observer Bundle — Reproducible verification artifact (observer-bundle-*.zip)
  • Receipts — Merkle-signed proof of every action
  • Ledger — Append-only causal log (.sxk-wal/)
  • Conformance — CRK-2 test suite (pnpm test:conformance)
# Verify a session independently
pnpm observer:replay --bundle observer-bundle-mission-002.zip
pnpm receipts:verify --session <session-id> --anchor sha256:...

See Evidence Hierarchy for the full proof model.


Project Structure

agentic-coding-agent/
├── agent/              # Agent implementations
├── backend/            # API server (gRPC/REST)
├── cockpit/            # Observability UI
├── control-tower/      # Multi-agent orchestration
├── crk2/               # Constitutional Runtime Kernel v2
├── docs/               # Architecture, guides, specs
├── src/
│   ├── kernel/         # Kernel core (invariants, receipts, lineage)
│   ├── runtime/        # Agent loops, tool execution
│   ├── persistence/    # WAL, snapshots, CSR
│   └── tools/          # File, Git, LSP, Exec, Web, Skills
├── config/             # Constitution, policies, schemas
├── examples/           # Example plans & workspaces
├── tests/              # Unit, integration, conformance
└── scripts/            # Build, deploy, verification

Contributing

All contributions require signed constitutional receipts:

# 1. Create a governed plan
pnpm contrib:plan --issue <number> --output plan.yaml

# 2. Run with governance
pnpm agent:run --plan plan.yaml --workspace . --sign-off

# 3. Submit with receipts
pnpm contrib:submit --plan plan.yaml --receipts receipts/

See CONTRIBUTING.md and CODE_OF_CONDUCT.md.


License

MIT with Constitutional Governance Addendum — see LICENSE.


Constitutional Profile

Property Value
Purpose Governed agentic coding & verification
Authority CRK-2 law kernel, Observer bundle, Control Tower
Evidence Model Receipts, ledger entries, observer bundle, docs
Verification Build, test, cockpit smoke, observer reproduction
Maturity Prototype (Mission #002)
Scope Lawful agentic coding, control tower coordination, cockpit visibility, observer verification
Limits Does not guarantee every production deployment or enterprise workflow

Governing Claim Rule: No repository should claim more than its evidence supports.


Blindspots & Battle Scars

  • Architectural: Repo is large; some paths still evolving
  • Governance: Some claims stronger than fresh evidence
  • Replay: Fresh verification needed across full surface
  • Operational: Not every launcher path refreshed this pass
  • Adoption: Repo can overwhelm without a scorecard

See Mission #002 for full maturity progression and validation ladder.

About

agentic coding agent

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Sponsor this project

Packages

Used by

Contributors

Languages