When teams ask what tools power an OpenClaw agent, it is a fair question. When someone tells you an AI runs a company's operations, you want to know what it is working with.

A typical OpenClaw tool stack grows organically over time. Some integrations are obvious from day one (email, calendar). Others emerge from specific needs (why a voice synthesis API? Because sometimes content is better heard than read). A few turn out to be dead ends — and this guide covers those honestly too.

Here is the complete map.

Communication

Gmail (Google Workspace)

What it does: Read, send, organize email How it is used: Morning inbox triage, customer response drafting, follow-up tracking Setup: OAuth 2.0 via Google Cloud Console Verdict: Essential. Email is the connective tissue of business.

Gmail is typically the first integration and remains the most impactful. The ability to read incoming emails and draft responses transforms an agent from "AI that gives advice" to "AI that handles things." For a full walkthrough, see How to Connect OpenClaw to Gmail.

Key configuration: Running with send_requires_approval: true for most contexts is recommended. Your team lead reviews important emails before they go out. For routine responses (meeting confirmations, receipt acknowledgments), the agent sends autonomously.

Slack

What it does: Real-time team communication How it is used: Team coordination, answering questions, posting updates Setup: Slack App with Socket Mode Verdict: Essential for teams. Less useful for solo founders.

A well-configured Slack presence is deliberately restrained. The agent responds to DMs, monitors key channels, and posts structured updates. It does not react to every message — that would be annoying. See the full OpenClaw + Slack setup guide for configuration details.

Technical detail: Socket Mode is important. It means no public webhooks, lower latency, and more reliable connections than the traditional Events API.

What it does: Direct messaging with your team How it is used: Primary 1:1 communication channel Setup: Bot token via BotFather Verdict: The daily driver.

Telegram is where many teams interact most with their agent. It is more personal than Slack, more immediate than email. Morning briefings, quick questions, approval requests, status updates — all through Telegram. For more detail, see the OpenClaw + Telegram guide.

Work Management

Linear

What it does: Issue tracking and project management How it is used: Creating issues, tracking progress, generating reports Setup: API key with GraphQL endpoint Verdict: Best project management tool for AI integration.

Linear's GraphQL API is a joy to work with. The agent can query anything — issues by team, by state, by assignee, by date range. It can create issues with full metadata: team, priority, labels, project, cycle.

Typical setup: Two teams (Development, Marketing), full state machine (Backlog, Todo, In Progress, In Review, Done), and a library of labels.

GitHub

What it does: Code hosting, version control, PRs How it is used: Pushing code, reviewing changes, managing repositories Setup: SSH key or personal access token Verdict: Essential for any agent that writes code.

An OpenClaw agent that writes code daily benefits enormously from GitHub access. It can commit, push, and create PRs without human intervention. The workflow: write code, test locally, commit with descriptive message, push, notify the team if review is needed.

Analytics & SEO

PostHog

What it does: Product analytics — pageviews, events, funnels How it is used: DAU tracking, feature usage, traffic analysis Setup: API key with project access Verdict: Great for web analytics. API is comprehensive.

PostHog provides the metrics that matter: daily active users, page-level analytics, event tracking, funnel analysis. The agent pulls these numbers for morning briefings and weekly reports.

One nice thing: PostHog's API lets you query with filters, so you can segment by traffic source, user property, or time range without pulling everything and filtering locally.

DataforSEO

What it does: SEO data — keyword rankings, volumes, competition How it is used: Weekly SEO reports, keyword research, competitor monitoring Setup: Basic auth (login/password) Verdict: Expensive but comprehensive. Best SEO API available.

DataforSEO is the engine behind automated weekly SEO reports. The relevant endpoints:

Keyword suggestions — discover new keywords related to a seed
Related keywords — find semantically connected terms
SERP analysis — check actual search result positions

Cost is per-request, so batching queries keeps spending reasonable. For more on setting up automated SEO reporting, see Automated SEO Reports Every Week.

Google Search Console (via Google Workspace)

What it does: Search performance data directly from Google How it is used: Click-through rates, impressions, indexation status Setup: Same OAuth as Gmail Verdict: The ground truth for SEO. Pairs perfectly with DataforSEO.

Social Media

PostBridge

What it does: Cross-platform social posting How it is used: Publishing to LinkedIn and Twitter from one API Setup: Bearer token authentication Verdict: Solves the multi-platform posting problem elegantly.

Connected accounts: 4 LinkedIn profiles/pages, 2 Twitter accounts. The agent drafts content adapted for each platform and schedules through a single API.

Apify

What it does: Web scraping actors (LinkedIn post search, social media data) How it is used: Monitoring competitors' social posts, researching trending content Setup: API key + LinkedIn cookies for authenticated scraping Verdict: Powerful but finicky. LinkedIn cookies expire frequently.

Apify is used primarily for LinkedIn content research — finding trending posts about AI agents, monitoring what competitors are publishing, and identifying content gaps.

Content & Voice

ElevenLabs

What it does: Text-to-speech with realistic voices How it is used: Converting blog posts to audio, voice narration Setup: API key Verdict: Surprisingly useful. Voice content adds a whole dimension.

This is often a late addition that teams do not expect to use much. It turns out that converting blog posts to audio creates a new consumption channel. Some people prefer listening to reading, especially for longer pieces.

Replicate

What it does: ML model inference — image generation, video, audio How it is used: Generating OG images, creating visual content Setup: API token Verdict: Good for specific tasks. Not a daily driver.

Replicate is used occasionally for generating images — blog post headers, social media graphics, OG images. The quality varies by model, but for functional images (not artistic ones), it is adequate.

Utility

Firecrawl

What it does: Web scraping and content extraction How it is used: Researching competitors, scraping documentation, gathering content Setup: API key Verdict: Clean API, reliable extraction. Used weekly.

When the agent needs to understand what a competitor is doing or gather information from a website that does not have an API, Firecrawl turns web pages into clean markdown. It is used for competitor content analysis and documentation research.

Brave Search

What it does: Web search How it is used: Real-time information lookup, research Setup: API key Verdict: Fast, private, good quality results.

Google Maps

What it does: Location services, geocoding, places How it is used: Weather data (paired with location), local business research Setup: API key with server IP restriction Verdict: Occasional use. Mostly for weather in morning briefings.

Gemini

What it does: Lightweight model inference How it is used: Quick, low-cost tasks that do not need Claude-level quality Setup: API key Verdict: Good for batch processing where quality matters less than cost.

The Dead Ends

Not everything works out. Here are the integrations that did not earn their keep:

DataFast

What it was supposed to do: Web analytics What it actually does: Visitor-level lookup and goal tracking. No aggregate analytics, no traffic data, no pageviews. Verdict: NOT useful for SEO reports or traffic monitoring. PostHog + GSC + DataforSEO covers everything DataFast does not.

Lesson learned: Always test the API endpoints before committing to a tool. The marketing page promised analytics; the API delivered visitor lookups.

How They Work Together

The real power is not in any single integration. It is in the combinations:

Email to Linear to Slack: Customer reports a bug via email, the agent creates a Linear issue, then notifies the team in Slack.

DataforSEO to Blog to PostBridge: The agent discovers a keyword opportunity, writes a blog post targeting it, then promotes it on social media.

PostHog to Telegram: The agent notices a metrics anomaly and proactively alerts the team lead.

Firecrawl to Content to PostBridge: The agent researches competitor content, identifies gaps, then writes and publishes to fill them.

Each integration is a node in a graph. The edges — the workflows connecting them — are where the value compounds. This is exactly the kind of system Toronto AI Consulting helps businesses build.

Adding New Integrations

When evaluating a new tool, consider these criteria:

Does it have an API? No API = no integration.
Does the API actually do what you need? (See: DataFast)
Does it connect to existing workflows? Isolated tools are less valuable.
Is the cost justified? Each API call costs money.
Can it be documented in a skill? If the API is too complex or undocumented, integration becomes fragile.

If a tool passes all five, add it to TOOLS.md with the API details and start building workflows around it. For a deeper look at the protocol that makes these integrations possible, see What Is MCP? A Business Guide to Model Context Protocol.

Current Configuration Size

For the curious:

Total integrations: 14 active
API keys managed: 11
OAuth connections: 2 (Google, Slack)
Cron jobs running: 4 (morning briefing, SEO report, social calendar, memory maintenance)
Skills installed: 6

This grows over time. Most teams add 1-2 new integrations per month as needs arise. The workspace evolves.