TENEX + OpenClaw

The Setup

You're using OpenClaw. It works. You don't want to migrate your entire setup.

TENEX runs alongside OpenClaw as an integration layer, not a replacement.

Five integration vectors. Each solves a specific gap. Each is honest about availability.

MCP Bridge

The Problem

OpenClaw has no MCP support. Requests for MCP integration were closed as NOT_PLANNED. Community PRs were closed without merge.

The community is desperate for MCP. Community-built claw-mcp has 8,000+ installs. Users are installing MCP support themselves.

TENEX Provides

TENEX exposes MCP endpoints natively. Your OpenClaw instance connects via standard MCP protocol.

// OpenClaw connects to TENEX MCP server
{
  "mcpServers": {
    "tenex": {
      "command": "tenex-mcp",
      "args": ["--project", "my-project"]
    }
  }
}

// Your OpenClaw agents now have access to:
// - lesson_learn / lesson_get
// - report_write / report_read
// - delegate / delegate_crossproject
// - RAG collections

Availability: In development. Targeting Q2 2026.

Persistent Memory Layer

The Problem

OpenClaw memory is broken.

What's broken:
• Session-scoped SQLite-vec dies on restart
• Memory compaction silently deletes context
• memoryFlush is disabled to prevent data loss
• Users report 6GB+ memory leaks without recovery

Your agent doesn't remember you between sessions.

TENEX Provides

TENEX RAG as persistent backend for your OpenClaw sessions. Every conversation is indexed, retrievable, and context-aware across restarts.

OpenClaw session
    │
    ▼
TENEX RAG store  ← Persistent across sessions
    │
    ▼
Nostr relay      ← Signed, verifiable, portable

// Agent learns something today
lesson_learn({ title: "OAuth pattern", lesson: "..." });

// Agent recalls it next week — different session
// No re-prompting. No context window stuffing.

Availability: In development. Targeting Q2 2026.

Multi-Agent Orchestration Upgrade

The Problem

OpenClaw multi-agent is file-based coordination. Agents pass files in a shared directory. No conversation. No delegation chain. No context about what the requesting agent needs.

Google research: naive multi-agent coordination degrades performance 39–70% on sequential reasoning tasks. Independent agents amplify errors 17x.

TENEX Provides

Real conversation-based coordination. Delegation chains where each agent brings accumulated project knowledge. Intelligent decomposition that routes sequential work to focused agents and fans out parallelizable work.

// File passing (OpenClaw)
agent_a → shared/output.json → agent_b
// No context, no provenance, no memory

// Real conversation (TENEX)
agent_a → delegate({ recipient: agent_b, prompt: "..." })
// → Full conversation persisted
// → Delegation chain cryptographically linked
// → Both agents bring project knowledge
// → Results flow back with context

Availability: In development. Targeting Q2 2026.

Identity Verification Layer

The Problem

OpenClaw agents have no identity beyond config files. No signing. No verification. You can't prove who created an agent, who published a skill, or whether output was tampered with.

TENEX Provides

Cryptographic identity for your OpenClaw agents via TENEX. Every action gets a Nostr keypair, a signature, and verifiable provenance. Your existing OpenClaw agents gain identity without changing their code.

// OpenClaw agent output — unsigned
{
  content: "Found 3 bugs in auth module"
  // Who wrote this? No way to verify.
}

// Same output via TENEX — signed
{
  content: "Found 3 bugs in auth module",
  pubkey:  "npub1def456...",
  sig:     "a3f8c2..."  // ✓ Verified
}

Availability: In development. Targeting Q2 2026.

Project Scoping

The Problem

OpenClaw has no project concept. Everything is a single workspace directory. No isolation. No scoping. Every agent sees everything. Context bleeds between tasks.

TENEX Provides

Project containers for your OpenClaw workspace. Each project has its own agents, memory, reports, and conversation history. Cross-project delegation when you need it. Isolation by default.

// Before: one flat workspace
~/workspace/
  ├── auth-code/
  ├── frontend-code/
  └── api-code/
  // Every agent sees everything. No isolation.

// After: TENEX project scoping
Project: "auth-service"
  ├── Agents: security-auditor, backend-dev
  ├── Memory: 47 lessons, 12 reports
  └── Scope: auth-code/ only

Project: "frontend"
  ├── Agents: ui-designer, svelte-expert
  ├── Memory: 23 lessons, 8 reports
  └── Scope: frontend-code/ only

// Cross-project when you need it
delegate_crossproject({ ... });

Availability: In development. Targeting Q2 2026.

How It Fits Together

TENEX sits beneath your existing OpenClaw setup. It doesn't replace your agents — it gives them infrastructure they're missing.

┌─────────────────────────────────────┐
│   Your OpenClaw Instance            │
│   · Existing agents                 │
│   · Existing skills                 │
│   · Existing workflows              │
└──────────────┬──────────────────────┘
               │ MCP Bridge
               ▼
┌─────────────────────────────────────┐
│   TENEX Layer                       │
│   · Persistent memory (RAG)         │
│   · Cryptographic identity          │
│   · Project scoping                 │
│   · Cross-project coordination      │
│   · Delegation chains               │
└──────────────┬──────────────────────┘
               │ Nostr protocol (signed events)
               ▼
┌─────────────────────────────────────┐
│   Nostr Relays (federated)          │
│   · Event storage                   │
│   · Message bus                     │
│   · Public or private               │
└─────────────────────────────────────┘

Your OpenClaw agents keep working. TENEX adds the substrate — memory that persists, identity that verifies, coordination that doesn't degrade.

Don't Switch. Layer.

TENEX is open source and early-stage. The integration layer is in active development. We're building in public. If you're an OpenClaw user hitting these walls, you're exactly who we're building for.