OpenClaw Slack Block Kit

English | 한국어

Turn structured OpenClaw answers into Slack-native dashboards, tables, image cards, and other Block Kit layouts.

• Sends to the current Slack channel, DM, or thread automatically.
• Reuses the Slack connection already configured in OpenClaw—no token or channel ID is exposed to the model.
• Adds one Slack-only tool, slack_send_blocks, without turning every reply into a card.

Before / After

The same fictional sales report rendered as a normal reply and with Block Kit.

| Plain text reply | With slack_send_blocks | |:---:|:---:| | | |

Install

Requirements:

• OpenClaw 2026.7.1-2 or later
• A working Slack connection in OpenClaw

Install from ClawHub:

openclaw plugins install clawhub:openclaw-slack-block-kit

Or install directly from npm:

openclaw plugins install npm:openclaw-slack-block-kit

After installation, no plugin-specific Slack credentials, destination, or other configuration is required. The plugin reuses OpenClaw's existing Slack connection and the current conversation route.

Use it from Slack

Start in the Slack conversation where you want the result, then ask naturally:

Show today's sales as a Slack dashboard with a concise summary and comparison table in this thread.

The agent can choose Block Kit when the answer benefits from a structured layout. To request it explicitly, name the tool:

Use slack_send_blocks to show these candidates as image cards in the current thread.

The message is always sent back to the Slack channel, DM, or thread where the request started. The tool cannot redirect it to another destination. For a different channel or thread, use OpenClaw's core message tool.

After a successful send, the Block Kit message is normally the final answer, without a duplicate plain-text reply. Each message still includes fallback text for Slack notifications and accessibility.

What it supports

• Comparisons, ranked results, status summaries, multi-field records, tables, image cards, grouped sections, and data visualizations
• Known display blocks: section (including fields and an image accessory), header, context, divider, image, rich_text, table, and data_visualization; common structure, size, URL, and safety checks run locally, while Slack validates their detailed schema
• Warning-bearing passthrough for unknown message-surface blocks, with the same common local checks
• 1–10 messages per call, in order; 1–50 blocks per message
• Required fallback text of 1–4,000 characters per message
• OpenClaw's existing Slack authentication, outbound queue, hooks, and delivery receipts

This plugin is display-only. It does not support buttons, selects, inputs, or any other element that requires an action_id; nor does it handle modals, App Home, or file/video/call lifecycles.

For a portable layout across channels, use OpenClaw's core message tool with presentation; use this plugin when you specifically need raw Slack Block Kit.

Troubleshooting

The tool does not appear

Plugin installation registers the tool automatically. The steps below grant access only when the active tool policy or runtime hides it.

1. Start the request from Slack. The tool is intentionally hidden on other surfaces.

2. Plugin installs and updates require the Gateway to reload. A managed Gateway normally restarts automatically; otherwise run openclaw gateway restart once.

3. Inspect the loaded runtime and confirm that the Gateway RPC is healthy:

   openclaw plugins inspect slack-block-kit --runtime --json
   openclaw gateway status --deep --require-rpc

4. If the plugin reports loaded but the tool is missing, run openclaw dashboard, then open OpenClaw Dashboard → Agents → select your agent → Tools, enable slack_send_blocks, and select Save. This is commonly needed when local onboarding selected the coding tool profile, which excludes third-party plugin tools.

5. If the plugin is disabled, run openclaw plugins enable slack-block-kit.

6. If plugins.allow is configured, make sure it includes slack-block-kit. This is a plugin-loading gate, separate from the agent's tool policy.

Profiles such as coding, messaging, and minimal exclude third-party plugin tools. The Dashboard toggle normally writes agents.list[].tools.alsoAllow for the selected agent. If the Dashboard reports that the agent uses an explicit allowlist, update that agent's tools.allow in the Dashboard Config tab instead.

To grant the tool globally, edit the active OpenClaw configuration file—normally ~/.openclaw/openclaw.json. A CLI --profile <name> (state isolation, unrelated to tool profiles) or OPENCLAW_CONFIG_PATH can change this location; run openclaw config file to print the active path. Add the following top-level tools entry to that file:

{
  tools: {
    alsoAllow: ["slack_send_blocks"],
  },
}

To limit the grant to one agent, put the same alsoAllow entry under that agent's agents.list[].tools instead. If the chosen scope already has tools.allow, add the tool to that array; allow and alsoAllow cannot coexist in one scope.

Sandbox tool policy is a separate gate. Sandboxed agents also need the plugin id slack-block-kit in tools.sandbox.tools.alsoAllow (or the existing sandbox allow array). Use group:plugins instead only to allow every plugin tool. For agent-, provider-, or sandbox-specific policies, see OpenClaw tool profiles and policies.

A send fails

• INVALID_ARGUMENT: check that each item in messages[] contains both text and blocks.
• INVALID_BLOCK_KIT: follow error.issues[].path and fix the reported block, URL, nesting, duplicate ID, or interactive element.
• INVALID_ROUTE: start the request from a real Slack channel, DM, or thread.
• RUNTIME_CONFIG_UNAVAILABLE: check that the Gateway and plugin runtime are loaded, then retry.
• SLACK_API_ERROR (for example, message: "invalid_blocks"): compare the payload with Slack's current block reference or test it in Block Kit Builder.
• SLACK_RATE_LIMITED: wait retryAfter seconds when provided; otherwise back off briefly before retrying.

Advanced usage

The agent normally builds this payload for you. The public input envelope is:

{
  "messages": [
    {
      "text": "Fallback for accessibility and notifications",
      "blocks": [
        {
          "type": "section",
          "text": { "type": "mrkdwn", "text": "*Candidate 1*" },
          "accessory": {
            "type": "image",
            "image_url": "https://example.com/item.png",
            "alt_text": "Candidate image"
          }
        }
      ]
    }
  ],
  "validateOnly": false
}

text and blocks belong inside every messages[] item. Flat top-level fields, missing fields, extra fields, and wrong types are rejected.

validateOnly: true runs local structure, size, and safety checks without sending to Slack. It does not guarantee that Slack will accept every block combination.

Multi-message sends are not atomic. After a partial_failed result, retry only the failed indexes to avoid duplicating messages that already succeeded.

Security

• The plugin neither accepts nor stores a separate Slack token.
• The destination comes only from the trusted current Slack route.
• URL-bearing fields require valid https: URLs, and model-facing errors omit tokens, full payloads, and internal stacks.

Do not put secrets or sensitive signed URLs in fallback text, blocks, or image URLs.

Develop from source

Node.js and pnpm are required only for source development.

git clone https://github.com/demarlik01/openclaw-slack-block-kit.git
cd openclaw-slack-block-kit
pnpm install --frozen-lockfile
pnpm build
openclaw plugins install --link "$PWD"
openclaw plugins enable slack-block-kit

Run the release checks with:

pnpm typecheck
pnpm test
pnpm verify:completion
pnpm plugin:metadata-check
pnpm plugin:validate
npm pack --dry-run

References

• Architecture: English · 한국어
• Slack: Block Kit overview · Block reference · Block Kit Builder
• OpenClaw: Plugin installation · Tool profiles and policies · Building plugins

License

MIT