agent-orchestration
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
agent-orchestration
Multi-agent orchestration runtime with operator-aware autonomy. Six specialized agents collaborate through a typed artifact pipeline, with autonomy levels dynamically adjusted based on operator cognitive state.
Architecture
┌──────────────────────────────────────────────────────────────┐
│ MCP Runtime (41 RPC methods) │
│ Sessions · Artifacts · Worktrees · Agent Execution · Merge │
└──────────────────────────┬───────────────────────────────────┘
│
┌──────────────────────────┴───────────────────────────────────┐
│ Orchestration Pipeline │
│ POLYMATH → RESONANT → ARCHITECT → EXECUTOR → HORIZON │
│ (optional) │
└────────┬─────────────────────────────────┬───────────────────┘
│ │
┌────────┴────────┐ ┌────────┴────────┐
│ MIND Engine │ │ BASIS Engine │
│ Operator State │ │ Context Score │
│ Machine (FSM) │ │ Calculator │
│ │ │ │
│ fresh → mid → │ │ Domains × │
│ faded ↔ high_ │ │ Weights → │
│ pressure │ │ Caps/Penalties → │
│ │ │ CSS (0-100) │
│ → Autonomy │ │ → Tier A/B/C │
│ L0-L3 │ │ │
└─────────────────┘ └──────────────────┘How It Works
Task enters the pipeline via the MCP runtime
MIND engine evaluates operator state (session duration, decision count, error rate, fatigue signals) and assigns an autonomy level (L0–L3)
BASIS engine calculates a Context Sufficiency Score across weighted domains — determines whether enough context exists to proceed
Agents execute in sequence, each producing a typed artifact:
Agent | Role | Artifact |
Polymath | Divergent exploration, hypothesis generation | TaskMap |
Resonant | Evidence validation, risk assessment | EvidencePack |
Architect | Design specification from evidence | DesignSpec |
Executor | Implementation in isolated worktrees | PatchSet |
Horizon | Ship gating (5 gates), release decision | ShipDecision |
Alchemist | Creative reframing when work is stuck | OptionsSet |
Permission matrix enforces read/write guardrails per agent per autonomy level. Only Executor can write code, and only at L2+.
Project Structure
slate/
├── agents/ # 6 agent runners (polymath, resonant, architect, executor, horizon, alchemist)
├── engines/
│ ├── basis/ # Context Sufficiency Score calculator
│ └── mind/ # Operator state machine + autonomy resolver
├── libs/
│ ├── context-schema/ # Zod artifact schemas (TaskMap, EvidencePack, DesignSpec, etc.)
│ ├── context-loader/ # Markdown + YAML context loading
│ ├── orchestration/ # Pipeline coordination + session management
│ └── llm-provider/ # LLM router (OpenAI primary, Anthropic fallback)
├── apps/
│ └── mcp-runtime/ # JSON-RPC server (41 methods)
├── tests/ # E2E test suites
└── tools/ # Agent manifests, contracts, templatesTech Stack
Language: TypeScript 5.4
Build: Nx monorepo
Schema Validation: Zod
Testing: Vitest
LLM Integration: OpenAI SDK, Anthropic SDK
Runtime: Node.js, vite-node
Protocol: MCP (Model Context Protocol) via JSON-RPC
Run
# Install dependencies
npm install
# Build all packages
npx nx run-many --target=build
# Run tests
npx nx run-many --target=test
# Start MCP runtime
npx nx run slate-runtime:serveStatus
Working prototype, actively developed. Core pipeline, all 6 agents, both engines, and the MCP runtime are functional with passing test suites. Not production-hardened — session persistence is file-based, no monitoring/observability layer yet.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Autonomous and governed SRE & MLOps control plane for AI agent platforms.
Multi-agent governance: task orchestration, compliance, decision validation, and ML predictions.
Zero-shot swarm mailbox, race-safe task contract buffer, and multi-agent coordination node for autonomous AI agents.
Governed autonomous-agent platform: HITL approvals, budget gates, audit chain, effect ledger.
