For the complete documentation index, see llms.txt. Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.
Attensira Logo
Attensira
Attensira for Agencies

Hand the client a
pull request, not a deck.

For agencies whose clients ship their sites from Git: one workspace per client, evidence you can quote verbatim from ChatGPT, Perplexity, Google AI, Claude and Gemini, and edits that arrive in the client's repo for their team to merge.

Get StartedBook A Demo
•Cancel anytime•One workspace per client
Platform

One workspace per client, one write path

16%Mention rate · n = 45

Every rate carries its sample size

Read it as 7 of 45 successful runs, not as a score. A change is only called a change once it clears a two-proportion z-test at 95%.

✦
Client workspace
WorkspaceAcme Corp
VoiceAcme brand profile
OutputPR → acme/website

Each client's voice, each client's repo

The brand profile and Skills for a workspace govern how the agent writes. The GitHub App is installed in the client's own org, scoped to the repositories they choose.

Attensira
OpenAI
Perplexity
Claude
Gemini

Drive it from your own stack

An MCP server with OAuth and a REST API keyed per workspace, so your own agents and reporting can read the same numbers. Open infrastructure, not a closed dashboard.

What a client review actually contains

One workspace per client, no blended number

Each client's prompts, answers, competitors and connections live in their own workspace, and an API key or MCP connection names exactly one. Every rate is reported with the count of successful runs behind it, so a client review starts from 7 of 45, not from a score.

See how tracking works
Mention rate by workspace · examplelast 7 days
Acme Corp
38%n = 45
TechFlow
20%n = 45
GreenLeaf
13%n = 15
DataPulse
0%n = 15
DataPulse: measured zero, not missing data
1:1
One workspace per client, no shared view
5
AI answer surfaces read on a schedule
CSV
Every table and chart exports its rows
PR
How work lands in a client's repo
Built for Agencies

What you can defend in a client review

Evidence with sample sizes, work that lands as a diff, and a plain account of what the numbers cannot say.

Report numbers that survive a client's scrutiny

Mention rate and citation rate, each with the count of successful runs behind it, and a change reported only when it clears a two-proportion z-test at 95%. When it does not, the report says the change is not provable rather than drawing a line.

See how tracking works

Show the client who is being named instead

Add competitors by domain and they are scored over the answers already collected, on the same denominator — no extra runs, no slots spent. Every cited domain is bucketed as own, competitor, AI platform, reference or third party; third party is where a citation is winnable.

See competitor tracking

Deliver the edit, not a slide about the edit

The agent writes against each client's brand profile and Skills and opens a pull request against their repository, naming the page and the finding in the body. Their engineers review and merge; you are not shipping a deck of recommendations.

See how the agent writes

Say plainly what it cannot do

There is no attribution: Attensira cannot connect a mention to a visit or a signup, and never claims to. Crawler fetches are a separate pipeline from citations. A client who hears that first trusts the rest of the number.

Talk to us

Where agency work goes next

The surfaces a client engagement runs on, and the sectors clients most often arrive from.

The agent, and what it is allowed to doThe inbox where every change waits for a yesCompetitor tracking across the tracked answer surfacesAI visibility for professional servicesHow to build a defensible GEO reportAttensira compared with the AI-visibility trackers

Start with one client
and one repository.

Track their buyer questions, read what the models answer today, and send the first pull request to their engineers.

Get StartedBook A Demo
•Cancel anytime•One workspace per client
FAQ

Frequently Asked Questions

Running this for several clients?
Book a walkthrough and we will set up the first workspace with you.
Book a demo

One workspace per client. Prompts, answers, competitors and connections belong to a workspace, and an API key or MCP connection names exactly one — which is why no MCP tool takes a workspace id. There is no cross-client view that mixes two clients' data into one number.

Prompt slots are a stock, not a monthly flow: a plan gives you a number of concurrently tracked prompts, and removing a prompt frees its slot immediately. The monthly response budget is the separate flow, consumed as readings are collected — prompts times models times sampling depth multiply into it. Both numbers per plan are on the pricing page.

Two things. The evidence: the verbatim answers, the mention and citation rates with the sample size behind each, and which domains got cited instead. And the work: pull requests opened against the client's own repository, which their engineers review and merge. Attensira never merges anything.

Every table and chart the agent draws carries Copy as TSV and Download as CSV of the underlying rows, and the MCP server and REST API let you pull the same numbers into your own stack. There is no white-label report product — what you get is the data, in your format.

Attribution. Attensira cannot tell you that a mention produced a visit, a citation produced a signup, or that either produced revenue — there is no identity join between a model's answer and a later visitor. Crawler fetches are counted separately from citations and the two are never reconciled. Say that up front and the rest of the reporting holds.