collimer

Name: collimer
Author: sandcastlelabs

What is Collimer?

CommunityOnlineNo sign-inGlobalFreeCan modify data

What it can do

List Primitives: List the granular tools available to this caller, with their covering intents. Use call_primitive to inspect or invoke one.
Call Primitive: Inspect or invoke a granular tool from list_primitives. Omit arguments or set describe to true to get its input schema and annotations; pass arguments to call it.
Get Audit: Read an audit's status, progress, and GEO score. Poll this after start_audit; read the finished plan with list_run_recommendations (via call_primitive).

What data it sees

Do you need an account

No: the server works without sign-in

What is Collimer?

Collimer is a remote MCP server for generative engine optimization (GEO). It measures how often AI answer engines — ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews — cite a brand, turns the gaps into a ranked fix plan, drafts the fixes for a person to review, and verifies whether what shipped moved the score. Trackers tell you you're invisible. Collimer gets you cited.

How to use Collimer?

Connect to the remote server at https://app.collimer.com/mcp.

Claude Code: claude mcp add --transport http collimer https://app.collimer.com/mcp
Claude Desktop: Settings → Connectors → Add custom connector, URL https://app.collimer.com/mcp
Cursor, VS Code, Windsurf and other remote-MCP clients: add {"type": "http", "url": "https://app.collimer.com/mcp"} under mcpServers.collimer.

There is no API key. Your client discovers the OAuth endpoints and asks you to sign in the first time it connects. A free Collimer account is enough to connect and run a scan.

Key features of Collimer

Scans a domain's AI-search visibility: a 0–100 score with a confidence interval, and the biggest gap.
Runs full audits and returns the finished report.
Lists open recommendations for a brand in plan-priority order.
Drafts the fix for a recommendation. Drafts go to a person for review; nothing is published automatically.
Verifies shipped work: checks whether a change landed and whether it moved the score.
Onboards a brand end to end, with proposed brand context a person confirms.
Remote Streamable HTTP with OAuth 2.1 (dynamic client registration, PKCE). Nothing to install.

Use cases of Collimer

Find out how often AI answer engines cite your brand, and where the biggest gap is.
Work a GEO fix plan from inside your coding or writing agent instead of a dashboard.
Draft the pages and edits that close a citation gap, with a human approving each one.
Confirm that published fixes moved the score, with the uncertainty stated.

FAQ from Collimer

What does a scan measure?

How often AI answer engines cite the domain for questions in its category. It returns a score with a confidence interval rather than a single point figure, plus the largest gap.

Do I need an API key?

No. The server uses OAuth: your MCP client prompts you to sign in on first connect. A free account works; drafting and some audit features depend on your plan.

Does Collimer read my files or conversation?

No. The server acts only on the Collimer account you sign in with and the brands and domains you ask it about.

Does it publish anything?

No. Drafts are returned for a person to review and publish. Verification checks what you shipped; it does not change your site.

How to connect

How to connect collimer to Claude

The server address to paste into Claude settings:

https://collimer--sandcastlelabs.run.tools

Open Claude (claude.ai or the desktop app).
Go to Settings → Connectors.
Click Add custom connector.
Paste the server address copied below into Remote MCP server URL and click Add.
In a chat, click + → Connectors and switch the new connector on.

Custom connectors are available on Free, Pro, Max, Team and Enterprise plans (Free is limited to one). On Team and Enterprise an organization Owner adds the connector first under Organization settings → Connectors.

Authorization

After you click Add, a sign-in window for the service opens. Sign in with your own account and approve access. Claude never sees your password.

How to connect collimer to ChatGPT

The server address to paste into ChatGPT settings:

https://collimer--sandcastlelabs.run.tools

Open ChatGPT in a browser (chatgpt.com). A Plus, Pro, Business, Enterprise or Edu plan is required.
Turn on developer mode once: Settings → Apps → Advanced settings → Developer mode.
Open Settings → Connectors and click Create.
Fill in the form: Name (anything), Description (one line about what the service does), MCP server URL (copy it below).
Under Authentication choose OAuth if the service requires sign-in, otherwise None. Click Create.
In a new chat open + → Apps/Connectors and enable the connector.

OpenAI has renamed this section before (Connectors → Apps/Plugins). If the label differs, search settings for "developer mode". On Business/Enterprise workspaces an admin must allow custom connectors first.

Authorization

On first use ChatGPT opens the service's sign-in window. Sign in and approve access.

How to connect collimer to Cursor

The server address to paste into Cursor settings:

https://collimer--sandcastlelabs.run.tools

Fastest: click Open in Cursor below and confirm the prompt.

Manually:

In Cursor open Settings → Cursor Settings → MCP and click Add new global MCP server.
Paste the JSON copied below into ~/.cursor/mcp.json (per project: .cursor/mcp.json in the repo root).
Save the file. The server appears in the MCP list; authorize it there if asked.

Authorization

If the service needs sign-in, an authorize button appears next to the server in the MCP list.

Server tool list (20)

Raw names from tools/list. Only developers need these.

list_primitives	List the granular tools available to this caller, with their covering intents. Use call_primitive to inspect or invoke one.
call_primitive	Inspect or invoke a granular tool from list_primitives. Omit arguments or set describe to true to get its input schema and annotations; pass arguments to call it.
get_audit	Read an audit's status, progress, and GEO score. Poll this after start_audit; read the finished plan with list_run_recommendations (via call_primitive).
get_audit_artifact	Read the assembled AI Visibility Audit deliverable for a completed run — the same report the audit download page serves. format: "md" returns the Markdown inline as this tool's result text. format: "pdf" returns a resource_link to the themed PDF instead of rendering it here, fetched separately with resources/read on that link's uri.
get_run	Read a run of any type, including its prompts, citation summaries, and sentiment. A daily slice (run_type "slice") returns score null with not_scored_reason "measured by reads": the reads own its numbers.
list_runs	List runs across the credential's organization ceiling, newest first. Daily slices (run_type "slice") are omitted unless asked for by run_type: they are observation sources for the reads, never scored runs, and always carry score null.
list_changes	List covered work changes for the acting organization since a cursor. Cursors are bound to one organization and expire after 183 days. Call once without a cursor to start with no backlog; use a separate cursor after switching organizations.
whoami	Who this credential is: the user (if any), the org it is acting as, every org it could act as, its consented and effective scopes, whether writing is enabled, and the acting org's project limit. acting_org (and so writing_enabled/project_limit) is null whenever this credential can reach more than one org — whoami always reports the full ceiling rather than the consented-org default other tools apply. Call this before addressing an org or hitting a plan limit blind. Collimer also offers writing tools. Ask Collimer to enable writing for your organization; reconnect afterward to refresh the tool list (subject to granted scopes).
get_recommendation	Full evidence for one audit recommendation — issue, evidence, rewrite, effort, verification history (was it implemented, when, what did verification detect), and every article it produced, if any (articles — working a recommendation more than once can produce several). Get an id from list_work (via call_primitive) with kind=recommendation.
get_recap	Read the monthly exec recap for a project: the score delta since the last run and the recommendations shipped in that window. Paid-plan feature — refused for a free-plan org. Returns structured content plus a Markdown rendering of the same recap the project's Recap page shows.
list_brands	List every brand (project) this credential can act on, with each one's last activity and how many articles need attention. Call this before `resolve_brand` when the domain isn't already known.
resolve_brand	Resolve a domain to the brand (project) it belongs to: org, project, latest audit run, and an opaque project handle every other writing operation takes. Call `list_brands` first if the domain isn't already known.
get_brand	Read a brand's settings and business classification.
create_brand	Create a brand (project) so a domain has somewhere to work in. Idempotent: calling this again for a domain you already hold returns that brand rather than erroring.
onboard_brand	Create or find a brand, start its first scan, and propose grounded brand context for approval. Re-run after collimer_scan_status (via call_primitive) completes to derive context from the scanned pages. Check each proposal's support status; correct contradicted or unverified values with the customer before confirming. Nothing is written to context unless confirm is true with reviewed facets.
update_brand	Update supplied brand fields. business_type reclassifies future scans.
context_schema	Get the brand-context facet catalog: every kind of context Collimer can hold about a brand, why it matters, what reads it, and how much to write. Call this before get_brand_context or update_brand_context — it is the only place these rules are written down.
get_brand_context	Get a brand's assembled context card — voice, positioning, stack, and everything else Collimer knows about it — plus a catalog of what exists but wasn't included, with sizes, so you can ask for it by name via `include`. Call `context_schema` first to see what each kind means.
review_brand_context	Review a brand's missing, unconfirmed, and stale context. Proposes evidence-backed changes from stored page snapshots without scanning or saving anything. Check each proposal's support status and correct contradicted or unverified values with the customer before saving confirmed values with update_brand_context. Never record a value as customer-asserted on their behalf.
update_brand_context	Write any subset of a brand's context facets in one call — voice, positioning, stack, and more. A body should be an excerpt of the brand's own words, not a summary; call context_schema first for the excerpt-vs-summary rule and what each basis value means.