Once Geckoboard is connected, your AI assistant can request metrics from supported data sources in your Geckoboard account. You can start asking questions immediately.
The optional skills on this page provide the assistant with reusable instructions for three common tasks: determining what data is available, preparing a periodic summary, and investigating a metric that changed.
Component | What it does |
Geckoboard MCP | Gives the AI assistant read-only access to metrics and, where supported, underlying rows. |
A skill | Tells the assistant how to approach a particular type of question. It does not add data or permissions. |
If you are not sure where to begin, install the Geckoboard Orientation skill and ask it to show you what is available.
Download the skills
You can install any or all of these skills.
Geckoboard Orientation
Use this first if you do not know what is connected or what questions your account can answer. It checks the available connections and metric catalogs before suggesting questions.
Geckoboard Pulse
Use this for a daily, weekly, monthly, or quarterly summary. It focuses on metrics that moved outside their recent pattern, reports what remained steady, and separates completed-period reviews from in-progress pace comparisons.
Geckoboard Metric Investigation
Use this when a specific number has changed, and you want to understand where the movement is concentrated. It checks the comparison period, volume, available breakdowns, and underlying rows. It can identify associations and possible explanations, but it cannot prove cause from metrics alone.
Install a skill in ChatGPT
Personal Skills are generally available on ChatGPT Business, Enterprise, Healthcare, and Edu. Your workspace administrator controls whether you can create, upload, and install skills.
To install a skill:
Download the skill from this page.
In the ChatGPT sidebar, select Plugins.
In the Plugin Directory, select Skills.
Select Create, then Upload from your computer.
Choose the downloaded
.skillfile and complete the installation.Repeat these steps for any other Geckoboard skills you want to use.
ChatGPT scans an uploaded skill before making it available. It may ask you to review the skill or block it if the scan identifies a potential risk.
Installing a skill does not connect ChatGPT to Geckoboard. Connect the Geckoboard MCP separately before using these skills. Skills added in ChatGPT may also need to be installed separately on desktop, web, or mobile.
Install a skill in Claude
Skills in Claude require Code execution and file creation to be enabled. On a Team or Enterprise account, an organization owner may also need to enable Skills.
To install a skill for your own Claude account:
Download the skill from this page.
In Claude, go to Customize, then Skills.
Select +, then Create skill.
Select Upload a skill and choose the downloaded file.
Make sure the skill is enabled in your Skills list.
Repeat these steps for any other Geckoboard skills you want to use.
The downloads are ZIP-format skill packages. If your client only allows files ending in .zip, rename the file extension from .skill to .zip before uploading it.
On Team and Enterprise plans, organization owners can provision skills for other users instead of asking each person to upload them.
If your AI client does not support skills, you can still use Geckoboard MCP and adapt the starter questions below.
Start by finding out what you can ask
Try the Orientation skill first:
Use the Geckoboard Orientation skill to show me which data sources are connected and suggest five questions I can ask about them.
You can also ask about a particular area:
Use Geckoboard to show me what support data is available and suggest questions about volume, response time, satisfaction and SLAs.
Can Geckoboard answer questions about our advertising spend and conversions? Check the available metrics rather than guessing.
The assistant may ask you to choose a connection if your Geckoboard account contains more than one connection to the same service.
Prepare a periodic summary
The Pulse skill works best with a stable set of metrics and a clear reader. On the first run, let it propose a set for you to approve.
Use the Geckoboard Pulse skill to prepare a summary of last week for our support lead. Check what is connected, then propose five to eight metrics before running the report.
If you already know the metrics you want:
Give me a pulse for last week using tickets created, tickets solved, first reply time, SLA attainment, and CSAT. Use the Europe/London timezone and compare the results with recent complete weeks.
For a period that is still in progress:
How is this month tracking so far? Compare it with the same elapsed portion of last month and describe it as a pace comparison.
The MCP supplies data when requested. Running the same pulse automatically each week depends on scheduling features in your AI client or workflow tool.
Investigate a change
Use the Metric Investigation skill when you have a particular movement to examine:
Use the Geckoboard Metric Investigation skill to investigate why Zendesk's first reply time increased last week. Confirm that the change is real, use an appropriate comparison period, check volume, and show where the movement is concentrated.
Investigate the fall in win rate last month. Check the underlying deal volume, compare relevant segments, and inspect a small set of underlying rows where available. Separate what the data shows from possible explanations.
Our call waiting time looks unusually high. Check whether this is outside recent observed variation and whether it is concentrated in a particular team or period.
An investigation may show where a change occurred and which other values moved alongside it. Confirming the operational cause may require information from the source system, such as audit history, staffing changes, or incident records.
Include enough detail in your question
You will usually get a better answer if you include:
the data source or connection, especially if you have connected the same service more than once;
the metric you want;
the timeframe and timezone;
any team, channel, region, or other filter;
the comparison you want;
who the answer is for or what decision it should inform.
You do not need to know the exact Geckoboard metric name. The assistant can search the catalog, but it may ask you to choose between similar definitions such as first reply time and resolution time.
What to expect
Geckoboard MCP is read-only. It cannot edit your source data, tickets, dashboards, or account settings.
The skills do not connect to additional data sources or give the assistant additional permissions.
Not every metric supports every timeframe, filter, or breakdown. The assistant should check what the selected metric allows.
Charts depend on the AI client's capabilities. A table and a written explanation can contain the same result.
Scheduling, sending messages, and updating documents depend on the AI or workflow tool, not solely on the MCP connection.
The AI assistant writes the interpretation. Review its timeframe, filters, and assumptions before relying on a consequential conclusion.
Metric calculations use the applicable data available to Geckoboard, but limits apply to the number of individual records or breakdown categories returned to the assistant. See Understanding result sizes in Geckoboard MCP.
Troubleshooting
I cannot see Skills in Claude
Check that Code execution and file creation are enabled. On Team or Enterprise accounts, ask an organization owner whether Skills are enabled.
I cannot see Skills in ChatGPT
Personal Skills are generally available on ChatGPT Business, Enterprise, Healthcare, and Edu. Ask your workspace administrator whether skills, skill installation, and skill uploads are enabled for your role.
The assistant is not using the skill
Check that the skill is installed and enabled. In ChatGPT, you can select or @ mention the skill. In Claude, check Customize → Skills. You can also name it directly in your request, for example: “Use the Geckoboard Orientation skill.”
The assistant cannot find my data
Check that:
the source is connected to Geckoboard, and its initial import has finished;
the source is supported by Geckoboard MCP;
the Geckoboard connector is enabled in your AI client;
you selected the correct connection when more than one is available.
The answer differs from a dashboard widget
Ask the assistant to state the connection, metric, aggregation, timeframe, timezone, filters, and breakdown it used. Compare those settings with the widget.