scanned Apr 20, 2026

getclaw

getclaw.sh

getclaw is a managed platform for deploying AI assistants into messaging channels

/100

Content answers—

Protocol plumbing—

Scored by asking 15 questions a buyer of a ai-ml product asks, then grading this site’s own pages: answered, hedged (partial or vague), or silent (no page answers it). How scoring works

15 buyer questions · graded against the site’s own pages

3 of 15 come back answered. 5 get a hedge. 7 meet silence.

3 answerable5 half-answers7 silent

Goes silent · api · high importance

We're building a multi-tenant setup where each customer gets their own AI assistant instance. What's the hard limit on concurrent API connections per account, and does that scale with your Team plan or require Enterprise?

Goes silent · api · medium importance

We're building billing logic on top of your webhooks for assistant usage. Do you guarantee at-least-once delivery with idempotency keys, or should we expect and handle duplicate webhook events?

This report is public. Own getclaw.sh? Claiming is free: crawl every page, re-audit as you fix, and track your score over time.

The fix queue

12 open questions, ordered by estimated payoff.

Point estimates are per fix under scoring v2. They are not additive to a promised total.

01api · importance highGoes silent+7 content pts est.

We're building a multi-tenant setup where each customer gets their own AI assistant instance. What's the hard limit on concurrent API connections per account, and does that scale with your Team plan or require Enterprise?

What the pages say

No page on the site addresses this.

The fix

Create a dedicated API limits or rate limits documentation page that specifies hard limits on concurrent API connections per account, with clear differentiation between Team and Enterprise plans. Include specific numbers (e.g., '100 concurrent connections on Team, 1000+ on Enterprise') and whether limits scale with plan tier or require custom negotiation.

confidence high · Absent

02api · importance mediumGoes silent+7 content pts est.

We're building billing logic on top of your webhooks for assistant usage. Do you guarantee at-least-once delivery with idempotency keys, or should we expect and handle duplicate webhook events?

What the pages say

No page on the site addresses this.

The fix

Create a dedicated webhooks documentation page that explicitly states the delivery guarantees (at-least-once, at-most-once, or exactly-once), whether idempotency keys are provided, and best practices for handling duplicate events. The current changelog mentions 'structured webhook delivery' for email but provides no technical details on reliability guarantees or deduplication mechanisms.

confidence high · Absent

03technical · importance mediumGoes silent+7 content pts est.

The Hermes v0.7.0 blog post mentions 'Memory Plugins' but doesn't specify the default retention window. If I don't configure anything, how long does conversation context persist before automatic purging?

What the pages say

No page on the site addresses this.

The fix

Add documentation for Memory Plugins default retention window in the Hermes v0.7.0 blog post or create a dedicated configuration reference page that specifies the default purging behavior when no retention is configured.

confidence high · Absent

04integration · importance mediumGoes silent+7 content pts est.

Our Discord community has aggressive auto-archive settings (24 hours). When Claw creates a thread for an AI assistant conversation, does it respect Discord's native archiving or force-keep threads open indefinitely?

What the pages say

No page on the site addresses this.

The fix

Create documentation specifically addressing Discord thread behavior, including whether Claw respects native auto-archive settings or implements any thread management to keep conversations accessible.

confidence high · Absent

05integration · importance mediumGoes silent+7 content pts est.

Telegram has granular bot permissions (message deletion, user restriction, etc.). Does Claw require full admin rights or can we scope permissions to only what's needed for conversational AI functionality?

What the pages say

No page on the site addresses this.

The fix

Create documentation specifically addressing Telegram bot permission requirements for Claw, including whether full admin rights are needed or if granular permissions can be scoped to conversational AI functionality only.

confidence high · Absent

06support · importance mediumGoes silent+7 content pts est.

We're evaluating you against a competitor who guarantees 4-hour response on critical outages. What's your actual SLA for production-impacting issues on the Team plan, and is there a phone escalation path?

What the pages say

No page on the site addresses this.

The fix

Create a dedicated support/SLA page that specifies response time commitments by plan tier (Team, Enterprise, etc.) and clearly documents escalation paths including phone support availability.

confidence high · Page missing

07limits · importance lowGoes silent+7 content pts est.

We're a services firm that wants to spin up separate AI assistants per client project. Is there a hard cap on total assistant instances per account, or can we programmatically create hundreds if each stays under message limits?

What the pages say

No page on the site addresses this.

The fix

Create a dedicated limits/pricing page or FAQ section that explicitly states whether there are hard caps on assistant instances per account, and whether programmatic creation of hundreds of assistants is supported.

confidence high · Absent

Hedged · 5 of 15 questions

A buyer gets something, then has to guess the rest. Expand any row for the evidence and the fix.

08Your security comparison mentions 'Credential Pools' in Hermes. If I have two separate AI assistants handling financial data versus general support, are their credential pools physically isolated or just logically namespaced?securityHedges+7 pts est.

Your security comparison mentions 'Credential Pools' in Hermes. If I have two separate AI assistants handling financial data versus general support, are their credential pools physically isolated or just logically namespaced?

What the pages say

No page on the site addresses this.

The fix

The chunks mention 'Credential Pools' as a v0.7.0 feature for automatic rotation of same-provider keys, but never explain whether pools are physically isolated (separate infrastructure) or logically namespaced (same infrastructure, separate labels). Add explicit documentation on credential pool isolation architecture, especially for multi-tenant or sensitive data separation use cases.

Where we looked: /blog/hermes-v0-7-0-founder-guide-memory-plugins-credential-pools, /blog/hermes-v0-7-0-founder-guide-memory-plugins-credential-pools, /blog/openclaw-vs-hermes-agent-security-model-comparison-2026, /blog/openclaw-vs-hermes-agent-security-model-comparison-2026

confidence medium · Vague

09Hermes v0.8.0 added background tasks. What's the maximum execution duration before a background task gets killed, and do failed tasks retry automatically or require manual intervention?technicalHedges+7 pts est.

Hermes v0.8.0 added background tasks. What's the maximum execution duration before a background task gets killed, and do failed tasks retry automatically or require manual intervention?

What the pages say

No page on the site addresses this.

The fix

The blog post about v0.8.0 mentions background tasks exist but provides no technical specifications. Add a dedicated technical documentation page or FAQ section covering: maximum execution duration/timeout limits, retry policies (automatic vs manual), and failure handling behavior for background tasks.

Where we looked: /blog/hermes-v0-8-0-founder-guide-background-tasks-live-model-switching

confidence high · Vague

10Your blog claims to serve 'non-technical founders' but the docs landing page is all API references. Is there a no-code path to deploy a basic Slack assistant, or do we need a developer for even the simplest setup?getting-startedHedges+2 pts est.

Your blog claims to serve 'non-technical founders' but the docs landing page is all API references. Is there a no-code path to deploy a basic Slack assistant, or do we need a developer for even the simplest setup?

What the pages say

getclaw supports deploying to Slack and claims you can deploy a fully configured AI assistant in under five minutes using a bring-your-own-key model. The changelog states Slack deployment uses a 'built-in setup guide' with 'step-by-step setup guides' and 'expandable guide with the exact steps, scopes, and permissions needed.' However, the actual docs/get-started page only shows API/curl commands with no visual UI-based path. The blog mentions 'If you want the fastest path, getclaw lets you deploy...to Telegram or your website in under five minutes' but specifically calls out Telegram for the tutorial, not Slack. For Slack specifically, the platform comparison describes it as 'The Enterprise Maze' with 'substantial bureaucratic overhead' and 'tedious' setup involving manual OAuth scopes, bot event subscriptions, and admin permissions. There is no explicit confirmation that a no-code UI path exists for Slack specifically—only that there are 'setup guides' of unspecified format.

The fix

Create a dedicated 'No-Code Slack Setup' page showing the actual dashboard UI steps, screenshots of the 'built-in setup guide,' and explicitly confirm whether Slack can be deployed without writing code or using API calls. Currently the 'setup guide' claim exists but its format (UI wizard vs. API docs) is unclear.

Where we looked: /blog/should-startup-hire-ai-employee-first, /changelog, /blog/telegram-slack-discord-ai-bot-comparison, /docs/get-started

confidence medium · Vague

11I'm running a customer support bot that handles roughly 50,000 messages per month across Slack and Telegram. At what exact message volume do your paid tiers kick in, and do you charge per message or per active conversation thread?pricingHedges+2 pts est.

I'm running a customer support bot that handles roughly 50,000 messages per month across Slack and Telegram. At what exact message volume do your paid tiers kick in, and do you charge per message or per active conversation thread?

What the pages say

getclaw charges per message, not per active conversation thread. The pricing model is: platform fee + $0.01-$0.05 per message in AI model API costs. At 2,000 messages/month, total cost is approximately $40-$125/month. However, the exact message volume threshold where paid tiers kick in is not specified—the chunks mention a 'Starter plan: $25/month' and 'Unlimited plan: $49/month' but do not state the message limits for each tier.

The fix

Add a clear pricing table showing exact message volume thresholds for each tier (e.g., 'Free: up to X messages/month', 'Starter: up to Y messages/month', 'Unlimited: unlimited messages'). Clarify whether the $25/month Starter plan includes any message credits or if all messages incur the $0.01-$0.05 API cost on top.

Where we looked: /faq, getclaw.sh, /assistants, /llms.txt

confidence medium · Vague

12The remote storage optimization post mentions Squashfs snapshots for Hermes agents. How frequently are these snapshots created by default, and can we trigger manual snapshots before risky deployments?technicalHedges+2 pts est.

The remote storage optimization post mentions Squashfs snapshots for Hermes agents. How frequently are these snapshots created by default, and can we trigger manual snapshots before risky deployments?

What the pages say

By default, Squashfs snapshots are created **every 15 minutes**. The blog post describes three automatic triggers: every 15 minutes, on container recreation (mount latest), and on SIGTERM (final snapshot before VM dies). However, the chunks do not explicitly state whether users can trigger **manual snapshots before risky deployments**. The post mentions that Hermes has since added a built-in `hermes backup` command with quick snapshot capabilities, but does not clarify if this is available for the Squashfs snapshot system described or only for general use.

The fix

Clarify whether the Squashfs snapshot system supports manual triggers (e.g., via API, CLI, or dashboard) for pre-deployment safety, or if users must rely solely on the 15-minute interval and SIGTERM capture.

Where we looked: /blog/hermes-remote-storage-squashfs-snapshots, /blog/hermes-remote-storage-squashfs-snapshots

confidence medium · Vague

What the pages already answer cleanly

3 of 15. Answered with enough specificity for a buyer to act on. This is the bar the queue above is chasing.

migrationAnswered

We're on OpenClaw now and considering the Hermes migration. The guide mentions 'execution speed' improvements but doesn't clarify: do we lose all existing conversation history and trained skill configurations during the switch?

What the pages say

Yes, you lose existing conversation history during the switch. The migration guide explicitly states: 'Export conversation history if you need it for reference. You will not import it into Hermes (the memory systems are different), but having a backup is good practice.' For trained skill configurations, you must manually recreate them: 'Seed initial skills. Hermes's self-improving system will build skills over time, but you can accelerate the process by manually creating skill documents for your most common tasks. Use your OpenClaw workspace skill files as templates.' The guide also notes that 'Inventory your OpenClaw skills. Which of the 52+ built-in skills do you use weekly? Export custom workspace skills as reference files' — indicating these skills do not automatically transfer and require manual recreation.

Source: /blog/switching-openclaw-to-hermes-agent-migration-guide-founders-2026, /blog/switching-openclaw-to-hermes-agent-migration-guide-founders-2026

confidence high · Answered

operationsAnswered

Your 'Why Open Source Matters' post promotes OpenClaw but doesn't specify the license. If we modify the core and deploy internally, are we obligated to publish those changes, or is there a commercial license exception?

What the pages say

OpenClaw is MIT-licensed, which means it is free to use, modify, and deploy commercially with no restrictions. There is no obligation to publish changes when modifying the core and deploying internally. The MIT license explicitly allows commercial use and modification without requiring source code disclosure.

Source: /blog/why-open-source-matters-ai-assistants-openclaw-primer

confidence high · Answered

technicalAnswered

The v0.8.0 'live model switching' feature sounds useful for A/B testing Claude vs GPT-4. Does switching models mid-conversation drop the active user session or maintain context seamlessly?

What the pages say

Live model switching maintains context seamlessly — it does not drop the active user session. According to the v0.8.0 founder guide, Hermes 'cuts that waste by letting the same task move between providers without losing thread context.' The feature specifically addresses the problem where 'switching means reconfiguring settings, opening a new session, or losing context' by making the switch 'part of the live workflow' within 'the same thread.'

Source: /blog/hermes-v0-8-0-founder-guide-background-tasks-live-model-switching

confidence high · Answered

Ask this site’s index

Sitedex already serves getclaw.sh as an MCP endpoint. Ask getclaw.sh anything an AI agent might ask, and see what its index returns. (To score your own site, use the form below.)

Snippets & configs

For developers and the engineer-on-call: copy these into your tools or your site.

Files from this audit

Built from this crawl. Download or copy each, then install it at the path noted.

llms.txt

Built from this crawl. Install at /llms.txt so agents start here.

organization.json

Organization JSON-LD, pre-filled from this crawl. Wrap in a ld+json script.

server-card.json

MCP server card built from this crawl. Host at /.well-known/mcp/server-card.json.

webmcp.json

WebMCP discovery manifest built from this crawl. Host at /.well-known/webmcp.json.

MCP endpoint

https://mcp.sitedex.dev/s/getclaw-sh/mcp

The URL anyone's agent points at. Read-only; safe to share.

Claude Code

claude mcp add getclaw --transport http https://mcp.sitedex.dev/s/getclaw-sh/mcp

One command, then the agent has it.

Cursor / Continue

{
  "mcpServers": {
    "getclaw": {
      "url": "https://mcp.sitedex.dev/s/getclaw-sh/mcp"
    }
  }
}

Drop into mcp.json.

WebMCP: two parts

WebMCP-capable browsers run the widget at runtime. Crawlers without JS rendering need the discovery manifest to find your tool surface. Install both.

1 · Widget script

<script async src="https://sitedex.dev/widget.js"></script>

Drop in <head>. WebMCP-capable browsers (Chrome 146+ Origin Trial) call navigator.modelContext.provideContext() via this script.

2 · Discovery manifest

{
  "$schema": "https://wellknownmcp.org/schemas/webmcp.json",
  "name": "getclaw.sh",
  "tools": [
    { "name": "search", "description": "Search getclaw.sh's indexed content." },
    { "name": "get_page", "description": "Fetch a page from getclaw.sh as markdown." }
  ]
}

Host alongside the script at /.well-known/webmcp.json. Crawlers that don't render JS rely on this.

Your turn

See which of these questions your site goes silent on.

Free, about 5 minutes. We crawl your site, test it against the buyer questions your category asks, and name what’s vague, contradictory, or missing, plus the files AI agents look for.

ComingEmbeddable grade badgeScore history and deltasOpt-in public board