How to Migrate to OpenClaw: From ClawdBot, MoltBot, or Other Tools [2026]

Why Migrate to OpenClaw?

If you are running ClawdBot, MoltBot, or another AI agent tool, you might be wondering whether it is worth migrating to OpenClaw. The short answer: yes, for most operators.

OpenClaw is the unified successor to both ClawdBot and MoltBot. As of early 2026, all active development has moved to the OpenClaw project. ClawdBot and MoltBot still work, but they are no longer receiving updates, security patches, or new skills. The community has consolidated around OpenClaw, which means support, documentation, and new integrations are all focused there.

Beyond the naming change, OpenClaw introduced significant architectural improvements in the 3.22 and 3.23 releases: a new plugin system, improved memory management, better multi-model routing, and native CLI commands for skill management. These improvements are not being backported to ClawdBot or MoltBot.

If you are running ZeroClaw or another tool entirely (AutoGPT, a custom LangChain agent, etc.), migration to OpenClaw makes sense if you want the messaging-first architecture, persistent memory, and the extensive ClawHub ecosystem. If your current tool is working well for a different use case, there may be no reason to switch.

Migrating from ClawdBot

ClawdBot to OpenClaw is the easiest migration path because OpenClaw is a direct fork of ClawdBot. The codebase, configuration format, and data structures are nearly identical.

Automated migration:

# Stop your ClawdBot instance
docker compose down

# Clone the OpenClaw repository
git clone https://github.com/openclaw/openclaw.git
cd openclaw

# Run the migration command pointing to your ClawdBot directory
npx openclaw migrate --from clawdbot --source /path/to/clawdbot

# This command will:
# - Copy and rename environment variables (CLAWDBOT_* → OPENCLAW_*)
# - Transfer memory files and conversation history
# - Update skill references to the new namespace
# - Generate a migration report

What transfers automatically:

• All environment variables (renamed from CLAWDBOT_* to OPENCLAW_*)
• Memory files (.md format) — transferred as-is
• Conversation history and session data
• Custom persona files
• Workflow configurations

What needs manual attention:

• Custom skills: If you wrote custom SKILL.md files, the migration tool updates the header namespace but you should verify the trigger commands still work.
• Webhook URLs: If external services send webhooks to your ClawdBot instance, you need to update those URLs to point to the new OpenClaw endpoint.
• Docker volume mounts: Your docker-compose.yml file needs updated volume paths. The migration command generates a new docker-compose.yml, but if you had custom modifications, you will need to merge them manually.

Migrating from MoltBot

MoltBot to OpenClaw requires slightly more work because MoltBot used a different environment variable naming convention and an older configuration format.

# Run the migration command
npx openclaw migrate --from moltbot --source /path/to/moltbot

# MoltBot-specific conversions:
# - MOLTBOT_API_KEY → OPENCLAW_MODEL_API_KEY
# - MOLTBOT_WA_TOKEN → OPENCLAW_WHATSAPP_TOKEN
# - MOLTBOT_TG_TOKEN → OPENCLAW_TELEGRAM_TOKEN
# - MOLTBOT_MEMORY_DIR → OPENCLAW_MEMORY_PATH
# - config.yaml → .env format conversion

MoltBot used YAML configuration files instead of .env files. The migration command converts these to the .env format that OpenClaw uses. However, if you had complex YAML structures (nested configurations, conditional blocks), you may need to flatten some settings manually.

Key differences to be aware of:

• Memory format: MoltBot stored memories in a proprietary binary format in older versions. If your MoltBot installation predates version 2.8, the migration tool converts these to .md files, but you should review the output for accuracy.
• Skill compatibility: MoltBot skills used a different trigger syntax. The migration tool attempts to convert these, but complex skills with multi-step workflows may need manual rewriting.
• Plugin API: MoltBot plugins are not compatible with OpenClaw. If you wrote custom MoltBot plugins (JavaScript), you will need to port them to the OpenClaw TypeScript plugin API.

Migrating from ZeroClaw

ZeroClaw provides its own migration command for moving to OpenClaw:

# From your ZeroClaw installation
zeroclaw migrate openclaw --output /path/to/new/openclaw

# This exports:
# - Configuration (TOML → .env conversion)
# - Conversation history (SQLite → OpenClaw format)
# - Memory files (already .md format, direct copy)
# - Integration configs (limited — see below)

The ZeroClaw-to-OpenClaw migration is a one-way process. ZeroClaw's Rust-based architecture is fundamentally different from OpenClaw's TypeScript architecture, so the migration focuses on data and configuration rather than code compatibility.

What transfers:

• Conversation history and memory files
• Basic configuration (model API keys, messaging tokens)
• User preferences and persona settings

What does not transfer:

• ZeroClaw plugins: Written in Rust, these cannot run in OpenClaw. You need to find equivalent OpenClaw skills or write new TypeScript plugins.
• Custom TOML configs: Advanced ZeroClaw configurations (rate limiting, custom routing rules) need to be recreated using OpenClaw's configuration system.
• Performance characteristics: ZeroClaw's sub-10ms startup and 5MB RAM footprint will not carry over. OpenClaw uses 800MB-1.5GB RAM and starts in 8-15 seconds. Plan your infrastructure accordingly.

Migrating from Other Tools

If you are coming from AutoGPT, a custom LangChain agent, CrewAI, or another framework, there is no automated migration path. You will need to set up OpenClaw from scratch and manually transfer what you can.

From AutoGPT:

• Export your AutoGPT workspace data and convert task logs to OpenClaw memory .md files
• AutoGPT plugins do not transfer — find equivalents on ClawHub or write custom skills
• Your AI model API keys can be reused (same providers, different env var names)

From LangChain-based agents:

• LangChain is a framework, not a deployed agent. Your custom code will not transfer directly.
• If you built tools in LangChain, you can often rewrite them as OpenClaw skills with similar logic
• Vector store data (if using Pinecone, Weaviate, etc.) can be connected to OpenClaw through plugins

From CrewAI:

• CrewAI agent definitions can inform your OpenClaw persona and multi-agent configuration
• Task definitions can be converted to OpenClaw workflow_auto.md entries
• CrewAI tools need to be replaced with OpenClaw skills

Environment Variable Changes

The most common stumbling block during migration is environment variables. Here is a complete reference for the variable name changes:

Important: The actual API key values do not change. You use the same Anthropic, OpenAI, or Google API keys. Only the environment variable names that hold them change. Do not regenerate your API keys — just rename the variables.

If you have custom environment variables (for plugins or custom integrations), the naming convention is to replace the old prefix with OPENCLAW_. For example, CLAWDBOT_CUSTOM_CRM_URL becomes OPENCLAW_CUSTOM_CRM_URL.

Data Portability

OpenClaw takes data portability seriously. Here is what you need to know about each data type:

Memory files: OpenClaw stores all agent memory as plain .md (Markdown) files. If your previous tool also used .md files (ClawdBot, ZeroClaw), they transfer directly with no conversion needed. Copy them to your OpenClaw memory directory and the agent will read them on the next startup.

Conversation history: OpenClaw stores conversation history in a SQLite database. The migration tools convert from other formats (ClawdBot's SQLite schema, MoltBot's JSON logs, ZeroClaw's SQLite schema) to OpenClaw's format. If you are migrating from a tool that stores conversations differently, you can write a simple script to insert records into OpenClaw's SQLite database.

Skills and workflows: SKILL.md files from ClawdBot transfer with minor namespace updates. Skills from other platforms need to be rewritten in the OpenClaw SKILL.md format. The format is straightforward — a Markdown file with YAML frontmatter that defines triggers, permissions, and the skill's instructions.

Persona files: Your agent's persona (personality, rules, behavior) transfers as a .md file. If your old tool used a different format for persona definition, you will need to rewrite it in OpenClaw's persona.md format.

Running in Parallel

For production deployments, you should run your old agent and OpenClaw in parallel during the transition period. Here is how:

# Run OpenClaw on a different port
OPENCLAW_PORT=3001 docker compose up -d

# Your old agent continues running on port 3000
# Test each integration one at a time on the OpenClaw instance

# Once verified, switch your reverse proxy / DNS to point to OpenClaw
# Monitor for 24-48 hours before shutting down the old agent

During the parallel period, only one agent should be connected to each messaging channel at a time. If both agents are connected to WhatsApp simultaneously, you will get duplicate responses. Migrate one channel at a time: start with a low-traffic channel (like Telegram or Discord), verify it works, then move to your primary channel (usually WhatsApp).

Post-Migration Checklist

After migration, run through this checklist to verify everything works:

• Send a test message on each connected messaging channel and verify the agent responds
• Check memory access: Ask the agent something it should know from previous conversations
• Test each integration: Trigger a calendar event, send a test email, query your CRM
• Verify skills: Run each custom skill and check the output
• Check logs: Run docker compose logs -f and look for errors or warnings
• Monitor resource usage: Verify RAM and CPU usage are within expected ranges
• Test workflow automations: If you have scheduled workflows (workflow_auto.md), verify they trigger correctly
• Update webhook URLs: Ensure all external services now point to your OpenClaw endpoint
• Update documentation: If you have internal docs referencing ClawdBot or MoltBot commands, update them
• Back up your old installation: Keep a backup of your old agent's data for at least 30 days in case you need to reference it

Migration is a one-time effort that pays off immediately. OpenClaw's active development, growing skill ecosystem, and consolidated community make it the clear choice for AI agent operators going forward. The migration tools handle 90% of the work — the remaining 10% is verification and testing.