Agents and MCPs - Subsig Docs

Docs

Agents and MCPs

Give Claude, or any MCP client, read access to your AI Visibility data so you can ask for it in plain language instead of clicking through the dashboard.

The Subsig MCP server gives Claude (or any MCP client) read access to your AI Visibility data: prompts, tracked brands, sentiment, citations, and generated reports. Ask for it in plain language instead of clicking through the dashboard.

Server address: https://api.subsig.com/mcp. Read-only. OAuth or a local API key, depending on your client.

Connect in 3 steps

Get a key. In the Subsig dashboard, go to Settings → API and generate a key. It looks like sk_live_… and is shown once.
Add the server. Point your client at https://api.subsig.com/mcp. Exact steps differ by client, see below.
Authorize once. Paste the key into the “Connect to Subsig” screen. All 13 tools show up right after.

Set up your client

Same server, same key. Only where you paste each one changes.

Claude.ai

Click your workspace name in the bottom-left corner, then Settings.
Go to Settings → Connectors.
Click Add custom connector at the bottom of the Connectors list.
Paste the server URL https://api.subsig.com/mcp, then click Add.
Click the new Subsig connector, then Connect. This opens the “Connect to Subsig” page in a new tab.
Open a second tab for app.subsig.com/dashboard/api-keys if you don’t have a key yet, generate one, and copy it.
Back on the Subsig tab, paste the key into the Subsig API Key field and click Connect.
Start a chat and toggle Subsig on under the tools icon, if it isn’t already.

Claude Code

One command, run from anywhere:

claude mcp add --transport http subsig https://api.subsig.com/mcp

The first tool call opens your browser to the same “Connect to Subsig” screen. Paste your sk_live_… key there once. Check it connected with:

claude mcp list

Claude Desktop

Open your config file and add the subsig block, then restart Claude Desktop.

Platform	Config file
macOS	`~/Library/Application Support/Claude/claude_desktop_config.json`
Windows	`%APPDATA%\Claude\claude_desktop_config.json`

{
  "mcpServers": {
    "subsig": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.subsig.com/mcp"]
    }
  }
}

On first use, the same “Connect to Subsig” screen opens in your browser. Paste your sk_live_… key there.

Available tools

You don’t call these directly. Ask in plain language and Claude picks the right one. Every tool maps 1:1 to a read endpoint on the Subsig Public API. Required arguments are listed below; everything else is optional. Full request and response detail is in MCP technical details.

All 13 tools are read-only. None spend credits or change tracked data.

Workspace

Tool	Required arguments	Description
`list_workspaces`	None	List every workspace this API key can access.

Prompts & topics

Tool	Required arguments	Description
`list_ai_visibility_topics`	None	List the topics prompts are grouped under, with active prompt counts.
`list_ai_visibility_prompts`	None	List tracked prompts, optionally filtered to one topic.
`get_ai_visibility_prompt_detail`	`prompt_id`	One prompt’s visibility, share of voice, and recent AI answers.

Brands & visibility

Tool	Required arguments	Description
`list_ai_visibility_brands`	None	List every tracked brand (own + competitors), ranked by AI visibility.
`get_ai_visibility_overview`	None	The account-level snapshot: your score, rank, and top competitors.
`get_ai_visibility_brand_detail`	`name`	Deep dive on one named brand, usually a competitor.
`get_ai_visibility_sentiment`	None	Positive / neutral / negative split when a brand is mentioned.
`get_ai_visibility_platforms`	None	Visibility broken out by AI provider.
`get_ai_visibility_citations`	None	Domains and URLs AI answers cite when they mention a brand.
`get_ai_visibility_status`	None	Whether today’s visibility checks have finished running yet.

Agent reports

Tool	Required arguments	Description
`list_agent_reports`	None	List already-generated agent reports (Root Cause Analysis, Prompts Discovery, Competitive Benchmark, AI Visibility Roadmap).
`get_agent_report`	`report_id`	Full findings from one report, plus signed PDF/DOCX links.

What the numbers mean

The definitions behind every metric a tool returns.

Metric	What it measures
AI visibility score	Share of eligible checks where the brand was mentioned, smoothed so a brand with only one or two checks can’t spike to a false 100%. Calculated as `mentioned ÷ (eligible + 5) × 100`.
Share of voice	The same mention count as a plain, unsmoothed share of every tracked brand’s mentions. Every brand’s slice sums to 100%, which is why it can read differently from the visibility score on a thin sample.
Average position	Mean rank within an answer when mentioned, pulled toward the account’s average when a brand’s own sample is thin. Null if the brand has never been mentioned.
Rank	Sorts every brand by visibility score, then share of voice as a tiebreaker, then average position. A never-mentioned brand always sorts last.
Sentiment score	A count-weighted average across every scored mention. Labeled Positive at ≥50% positive share, Negative at ≥40% negative share, Neutral otherwise.
Citation	One instance of an AI answer linking out to a domain or URL while discussing the tracked brand. Classified owned, competitor, UGC (Reddit, G2, Trustpilot), institutional (.gov / .edu / .mil), or reference.

AI platforms covered

Provider	Available from	Refresh
ChatGPT (OpenAI)	Free and up	Daily
Perplexity	Starter and up	Daily
Google AI Overviews	Starter and up	Daily

What you can ask

Things that take one conversational turn here and several dashboard screens otherwise. Every example below was run for real, not invented: same key, same tools, real accounts.

Ask	Tools used	Real result
“Compare AI visibility between our jira and Zendesk workspaces.”	`list_workspaces`, `get_ai_visibility_overview`	jira 58.9% visibility, Zendesk 20.4%.
“Summarize our latest agent report for the team.”	`list_agent_reports`, `get_agent_report`	A completed Root Cause Analysis: “Citations from usepylon.com dropped by 6 in the last week.”
“How’s GitHub doing, and who’s citing them?”	`get_ai_visibility_brand_detail`, `get_ai_visibility_citations`	41.4% visibility, rank #2; most-cited domain atlassian.com (129 citations).
“Has today’s visibility check finished?”	`get_ai_visibility_status`	Ready, 96 of 100 checks completed.

FAQ

Can it change anything in my account?

No. All 13 tools are read-only. Nothing here creates, edits, or deletes a prompt, competitor, report, or setting.

Does it work on the Free plan?

It connects fine, but every AI Visibility call is blocked. Free’s MCP quota is zero. Starter and up get real access, with a monthly request quota that scales by plan.

I manage more than one workspace. Which one does it use?

Your key has one primary workspace, used by default. Ask “which workspaces can you see” and the agent lists every workspace your key can reach; naming one after that switches the rest of the conversation to it.

Can it generate a new report?

No. It only reads reports already generated from the dashboard. Kick off a new one from the dashboard first.

How far back does the history go?

As far as your plan’s history entitlement: 6 months on Starter, 12 on Pro, 24 on Advanced. Asking further back just returns what’s available inside that window.

What happens when my connection expires?

A hosted Claude.ai connection is valid for 365 days, with no silent renewal past that. Remove and re-add the connector. A local key in Claude Code or Desktop never expires on its own; only revoking it in the dashboard cuts it off.

Something’s not working. What do I check first?

Reconnect if you’re getting an authorization error. The key may be revoked or your plan’s payment grace period may have lapsed. See MCP technical details for the full list of response codes and what each one means.

💡 Ask “which Subsig tools can you see?” right after connecting. If all 13 come back, the key and the server are both working.

Server address: https://api.subsig.com/mcp. Read-only. OAuth or a local API key, depending on your client.

Connect in 3 steps

Get a key. In the Subsig dashboard, go to Settings → API and generate a key. It looks like sk_live_… and is shown once.
Add the server. Point your client at https://api.subsig.com/mcp. Exact steps differ by client, see below.
Authorize once. Paste the key into the “Connect to Subsig” screen. All 13 tools show up right after.

Set up your client

Same server, same key. Only where you paste each one changes.

Claude.ai

Click your workspace name in the bottom-left corner, then Settings.
Go to Settings → Connectors.
Click Add custom connector at the bottom of the Connectors list.
Paste the server URL https://api.subsig.com/mcp, then click Add.
Click the new Subsig connector, then Connect. This opens the “Connect to Subsig” page in a new tab.
Open a second tab for app.subsig.com/dashboard/api-keys if you don’t have a key yet, generate one, and copy it.
Back on the Subsig tab, paste the key into the Subsig API Key field and click Connect.
Start a chat and toggle Subsig on under the tools icon, if it isn’t already.

Claude Code

One command, run from anywhere:

claude mcp add --transport http subsig https://api.subsig.com/mcp

The first tool call opens your browser to the same “Connect to Subsig” screen. Paste your sk_live_… key there once. Check it connected with:

claude mcp list

Claude Desktop

Open your config file and add the subsig block, then restart Claude Desktop.

Platform	Config file
macOS	`~/Library/Application Support/Claude/claude_desktop_config.json`
Windows	`%APPDATA%\Claude\claude_desktop_config.json`

{
  "mcpServers": {
    "subsig": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.subsig.com/mcp"]
    }
  }
}

On first use, the same “Connect to Subsig” screen opens in your browser. Paste your sk_live_… key there.

Available tools

All 13 tools are read-only. None spend credits or change tracked data.

Workspace

Tool	Required arguments	Description
`list_workspaces`	None	List every workspace this API key can access.

Prompts & topics

Tool	Required arguments	Description
`list_ai_visibility_topics`	None	List the topics prompts are grouped under, with active prompt counts.
`list_ai_visibility_prompts`	None	List tracked prompts, optionally filtered to one topic.
`get_ai_visibility_prompt_detail`	`prompt_id`	One prompt’s visibility, share of voice, and recent AI answers.

Brands & visibility

Tool	Required arguments	Description
`list_ai_visibility_brands`	None	List every tracked brand (own + competitors), ranked by AI visibility.
`get_ai_visibility_overview`	None	The account-level snapshot: your score, rank, and top competitors.
`get_ai_visibility_brand_detail`	`name`	Deep dive on one named brand, usually a competitor.
`get_ai_visibility_sentiment`	None	Positive / neutral / negative split when a brand is mentioned.
`get_ai_visibility_platforms`	None	Visibility broken out by AI provider.
`get_ai_visibility_citations`	None	Domains and URLs AI answers cite when they mention a brand.
`get_ai_visibility_status`	None	Whether today’s visibility checks have finished running yet.

Agent reports

Tool	Required arguments	Description
`list_agent_reports`	None	List already-generated agent reports (Root Cause Analysis, Prompts Discovery, Competitive Benchmark, AI Visibility Roadmap).
`get_agent_report`	`report_id`	Full findings from one report, plus signed PDF/DOCX links.

What the numbers mean

The definitions behind every metric a tool returns.

Metric	What it measures
AI visibility score	Share of eligible checks where the brand was mentioned, smoothed so a brand with only one or two checks can’t spike to a false 100%. Calculated as `mentioned ÷ (eligible + 5) × 100`.
Share of voice	The same mention count as a plain, unsmoothed share of every tracked brand’s mentions. Every brand’s slice sums to 100%, which is why it can read differently from the visibility score on a thin sample.
Average position	Mean rank within an answer when mentioned, pulled toward the account’s average when a brand’s own sample is thin. Null if the brand has never been mentioned.
Rank	Sorts every brand by visibility score, then share of voice as a tiebreaker, then average position. A never-mentioned brand always sorts last.
Sentiment score	A count-weighted average across every scored mention. Labeled Positive at ≥50% positive share, Negative at ≥40% negative share, Neutral otherwise.
Citation	One instance of an AI answer linking out to a domain or URL while discussing the tracked brand. Classified owned, competitor, UGC (Reddit, G2, Trustpilot), institutional (.gov / .edu / .mil), or reference.

AI platforms covered

Provider	Available from	Refresh
ChatGPT (OpenAI)	Free and up	Daily
Perplexity	Starter and up	Daily
Google AI Overviews	Starter and up	Daily

What you can ask

Things that take one conversational turn here and several dashboard screens otherwise. Every example below was run for real, not invented: same key, same tools, real accounts.

Ask	Tools used	Real result
“Compare AI visibility between our jira and Zendesk workspaces.”	`list_workspaces`, `get_ai_visibility_overview`	jira 58.9% visibility, Zendesk 20.4%.
“Summarize our latest agent report for the team.”	`list_agent_reports`, `get_agent_report`	A completed Root Cause Analysis: “Citations from usepylon.com dropped by 6 in the last week.”
“How’s GitHub doing, and who’s citing them?”	`get_ai_visibility_brand_detail`, `get_ai_visibility_citations`	41.4% visibility, rank #2; most-cited domain atlassian.com (129 citations).
“Has today’s visibility check finished?”	`get_ai_visibility_status`	Ready, 96 of 100 checks completed.

FAQ

Can it change anything in my account?

No. All 13 tools are read-only. Nothing here creates, edits, or deletes a prompt, competitor, report, or setting.

Does it work on the Free plan?

It connects fine, but every AI Visibility call is blocked. Free’s MCP quota is zero. Starter and up get real access, with a monthly request quota that scales by plan.

I manage more than one workspace. Which one does it use?

Can it generate a new report?

No. It only reads reports already generated from the dashboard. Kick off a new one from the dashboard first.

How far back does the history go?

As far as your plan’s history entitlement: 6 months on Starter, 12 on Pro, 24 on Advanced. Asking further back just returns what’s available inside that window.

What happens when my connection expires?

Something’s not working. What do I check first?

💡 Ask “which Subsig tools can you see?” right after connecting. If all 13 come back, the key and the server are both working.

Server address: https://api.subsig.com/mcp. Read-only. OAuth or a local API key, depending on your client.

Connect in 3 steps

Get a key. In the Subsig dashboard, go to Settings → API and generate a key. It looks like sk_live_… and is shown once.
Add the server. Point your client at https://api.subsig.com/mcp. Exact steps differ by client, see below.
Authorize once. Paste the key into the “Connect to Subsig” screen. All 13 tools show up right after.

Set up your client

Same server, same key. Only where you paste each one changes.

Claude.ai

Click your workspace name in the bottom-left corner, then Settings.
Go to Settings → Connectors.
Click Add custom connector at the bottom of the Connectors list.
Paste the server URL https://api.subsig.com/mcp, then click Add.
Click the new Subsig connector, then Connect. This opens the “Connect to Subsig” page in a new tab.
Open a second tab for app.subsig.com/dashboard/api-keys if you don’t have a key yet, generate one, and copy it.
Back on the Subsig tab, paste the key into the Subsig API Key field and click Connect.
Start a chat and toggle Subsig on under the tools icon, if it isn’t already.

Claude Code

One command, run from anywhere:

claude mcp add --transport http subsig https://api.subsig.com/mcp

The first tool call opens your browser to the same “Connect to Subsig” screen. Paste your sk_live_… key there once. Check it connected with:

claude mcp list

Claude Desktop

Open your config file and add the subsig block, then restart Claude Desktop.

Platform	Config file
macOS	`~/Library/Application Support/Claude/claude_desktop_config.json`
Windows	`%APPDATA%\Claude\claude_desktop_config.json`

{
  "mcpServers": {
    "subsig": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.subsig.com/mcp"]
    }
  }
}

On first use, the same “Connect to Subsig” screen opens in your browser. Paste your sk_live_… key there.

Available tools

All 13 tools are read-only. None spend credits or change tracked data.

Workspace

Tool	Required arguments	Description
`list_workspaces`	None	List every workspace this API key can access.

Prompts & topics

Tool	Required arguments	Description
`list_ai_visibility_topics`	None	List the topics prompts are grouped under, with active prompt counts.
`list_ai_visibility_prompts`	None	List tracked prompts, optionally filtered to one topic.
`get_ai_visibility_prompt_detail`	`prompt_id`	One prompt’s visibility, share of voice, and recent AI answers.

Brands & visibility

Tool	Required arguments	Description
`list_ai_visibility_brands`	None	List every tracked brand (own + competitors), ranked by AI visibility.
`get_ai_visibility_overview`	None	The account-level snapshot: your score, rank, and top competitors.
`get_ai_visibility_brand_detail`	`name`	Deep dive on one named brand, usually a competitor.
`get_ai_visibility_sentiment`	None	Positive / neutral / negative split when a brand is mentioned.
`get_ai_visibility_platforms`	None	Visibility broken out by AI provider.
`get_ai_visibility_citations`	None	Domains and URLs AI answers cite when they mention a brand.
`get_ai_visibility_status`	None	Whether today’s visibility checks have finished running yet.

Agent reports

Tool	Required arguments	Description
`list_agent_reports`	None	List already-generated agent reports (Root Cause Analysis, Prompts Discovery, Competitive Benchmark, AI Visibility Roadmap).
`get_agent_report`	`report_id`	Full findings from one report, plus signed PDF/DOCX links.

What the numbers mean

The definitions behind every metric a tool returns.

Metric	What it measures
AI visibility score	Share of eligible checks where the brand was mentioned, smoothed so a brand with only one or two checks can’t spike to a false 100%. Calculated as `mentioned ÷ (eligible + 5) × 100`.
Share of voice	The same mention count as a plain, unsmoothed share of every tracked brand’s mentions. Every brand’s slice sums to 100%, which is why it can read differently from the visibility score on a thin sample.
Average position	Mean rank within an answer when mentioned, pulled toward the account’s average when a brand’s own sample is thin. Null if the brand has never been mentioned.
Rank	Sorts every brand by visibility score, then share of voice as a tiebreaker, then average position. A never-mentioned brand always sorts last.
Sentiment score	A count-weighted average across every scored mention. Labeled Positive at ≥50% positive share, Negative at ≥40% negative share, Neutral otherwise.
Citation	One instance of an AI answer linking out to a domain or URL while discussing the tracked brand. Classified owned, competitor, UGC (Reddit, G2, Trustpilot), institutional (.gov / .edu / .mil), or reference.

AI platforms covered

Provider	Available from	Refresh
ChatGPT (OpenAI)	Free and up	Daily
Perplexity	Starter and up	Daily
Google AI Overviews	Starter and up	Daily

What you can ask

Things that take one conversational turn here and several dashboard screens otherwise. Every example below was run for real, not invented: same key, same tools, real accounts.

Ask	Tools used	Real result
“Compare AI visibility between our jira and Zendesk workspaces.”	`list_workspaces`, `get_ai_visibility_overview`	jira 58.9% visibility, Zendesk 20.4%.
“Summarize our latest agent report for the team.”	`list_agent_reports`, `get_agent_report`	A completed Root Cause Analysis: “Citations from usepylon.com dropped by 6 in the last week.”
“How’s GitHub doing, and who’s citing them?”	`get_ai_visibility_brand_detail`, `get_ai_visibility_citations`	41.4% visibility, rank #2; most-cited domain atlassian.com (129 citations).
“Has today’s visibility check finished?”	`get_ai_visibility_status`	Ready, 96 of 100 checks completed.

💡 Ask “which Subsig tools can you see?” right after connecting. If all 13 come back, the key and the server are both working.

Add Workspace

MCP technical details