AI Native Lang

OpenClaw integration

Hub (all MCP hosts): gettingstarted/HOSTMCPINTEGRATIONS.md(gettingstarted/HOSTMCPINTEGRATIONS.md) — ainl install-mcp --host openclaw (same as ainl install-openclaw).

OpenClaw integration

Hub (all MCP hosts): getting_started/HOST_MCP_INTEGRATIONS.md — ainl install-mcp --host openclaw (same as ainl install-openclaw).

PyPI: ainativelang v1.3.3.

OpenClaw Skill

AINL ships an OpenClaw skill under skills/openclaw/ (deterministic graphs, Markdown importer, ainl-mcp) and ainl install-mcp --host openclaw (alias ainl install-openclaw), a user-side bootstrap that wires PyPI, ~/.openclaw/openclaw.json (mcp.servers.ainl), and ~/.openclaw/bin/ainl-run without changing the OpenClaw application itself.

Standalone skill repo (optional, later): copy skills/openclaw/ to github.com/sbhooley/ainl-openclaw-skill as the repository root (SKILL.md, install.sh, README.md) if you want a single-purpose repo for ClawHub or docs links.

Why this matters: AINL is compile-once, run-many—you pay authoring or import cost once, then execute a validated graph repeatedly. Size economics use tiktoken cl100k_base; on the viable subset of representative workloads, minimal_emit lands near ~1.02× leverage vs unstructured baselines (see BENCHMARK.md and benchmarks.md for methodology and legacy-inclusive transparency).

OpenClaw normally uses npm install -g openclaw (or project-local install) and openclaw onboard. This skill adds the Python ainl toolchain and merges stdio ainl-mcp into the host MCP table. OpenClaw does not use ZeroClaw’s zeroclaw skills install <url> — prefer ClawHub (when listed) or manual copy into ~/.openclaw/skills or <workspace>/skills (default workspace is often ~/.openclaw/workspace).

Official OpenClaw source: use openclaw.ai for the installer and docs (e.g. curl -fsSL https://openclaw.ai/install.sh | bash). There is no stable public openclaw/openclaw GitHub repo advertised as the primary home.

Cron / bridge alternative: operator workflows that run AINL from OpenClaw cron (wrappers, token budget alerts, etc.) live under openclaw/bridge/ in this repo — see openclaw/bridge/README.md. That path is separate from this MCP skill.

Apollo X promoter (supervised HTTP gateway + ExecutorBridgeAdapter): the graph in apollo-x-bot/ainl-x-promoter.ainl expects a long-lived apollo-x-bot/gateway_server.py and ainl run with --enable-adapter bridge (plus memory). You must schedule polls separately: OpenClaw cron or OS cron should call apollo-x-bot/openclaw-poll.sh (or an equivalent command line). The graph’s S core cron line is documentation only and does not register a job. That script sets a long --http-timeout-s so batched llm.classify over the bridge does not abort at the default 5s client limit. Production layout, copy-paste openclaw cron add, and env files: apollo-x-bot/OPENCLAW_DEPLOY.md · operator notes: apollo-x-bot/README.md (scheduling + troubleshooting).

Memory surfaces: durable structured workflow state uses the SQLite memory adapter (see docs/adapters/MEMORY_CONTRACT.md). OpenClaw bridge cron may append daily markdown under ~/.openclaw/workspace/memory/, which is orthogonal to that adapter. ZeroClaw uses the same AINL memory/MCP path—not OpenClaw’s markdown layout. Narrative: AINL, structured memory, and OpenClaw-style agents.

CLI: ainl install-mcp --host openclaw (same as install-openclaw) ships in current ainl releases; upgrade from PyPI if your install reports an unknown command.

Quickstart

  1. Install the skill — copy skills/openclaw/ into your OpenClaw skills directory, or install via ClawHub when available, then:

    cd /path/to/skill && chmod +x install.sh && ./install.sh

    install.sh optionally refreshes the OpenClaw CLI via npm, upgrades ainl[mcp], and runs ainl install-mcp --host openclaw. To skip the global npm step: OPENCLAW_SKIP_NPM=1 ./install.sh.

  2. Or run the bootstrap directly (Python only):

    pip install 'ainativelang[mcp]'
ainl install-mcp --host openclaw

    Equivalent: ainl install-openclaw. Preview: --dry-run · noisy logs: --verbose.

Chat example

After a successful install, try:

Import the morning briefing using AINL.

Then use ainl import markdown …, ecosystem shortcuts (ainl import clawflows / ainl import agency-agents), or MCP tools so the agent produces compiling .ainl source, followed by ainl compile / ainl run or ~/.openclaw/bin/ainl-run my.ainl.

Optional adapter — code_context: graphs that call R code_context.* (tiered repo index, dependencies, impact, COMPRESS_CONTEXT) must pass --enable-adapter code_context to ainl run or ~/.openclaw/bin/ainl-run (extra args forward to ainl run). Installing MCP does not enable optional adapters. Guide: docs/adapters/CODE_CONTEXT.md · demo examples/code_context_demo.ainl · optional env AINL_CODE_CONTEXT_STORE.

What gets installed

| Artifact | Purpose | |----------|---------| | pip install --upgrade 'ainativelang[mcp]' | Latest compiler, importer extras, MCP dependencies | | ~/.openclaw/openclaw.json | Merges mcp.servers.ainl stdio entry pointing at ainl-mcp (skipped if already present with the same resolved command); other top-level keys preserved | | ~/.openclaw/bin/ainl-run | Shell wrapper: compile then exec ainl run with extra args forwarded | | ~/.bashrc / ~/.zshrc | Appends export PATH="$HOME/.openclaw/bin:$PATH" when those files exist and do not already mention ~/.openclaw/bin |

If no shell rc file is updated, the command prints a one-line PATH tip you can paste manually.

Benefits (summary)

| Benefit | Notes | |---------|--------| | Deterministic graphs | Compile-time validation; explicit cron, steps, and branches vs prose-only prompts | | OpenClaw-native paths | openclaw.json + ~/.openclaw/bin align with OpenClaw’s config layout | | MCP parity | Same import semantics as CLI; exposure profiles in tooling/mcp_exposure_profiles.json | | ~1.02× viable leverage | On tokenizer-aligned viable subset; see benchmarks for honest caveats |

See also

LLM Adapter Setup

To enable cloud LLM providers, create an AINL config file following the guide in docs/LLM_ADAPTER_USAGE.md. Set your provider API keys via environment variables (e.g., OPENROUTER_API_KEY). Then pass the config path using --config or set AINL_CONFIG.

You can verify cost tracking and retry behavior by running a program and checking the SQLite DB under intelligence/monitor/.