Flashpoint.AIFlashpoint.AIdocs

MCP capability reference

Everything an AI assistant connected to Flashpoint.AI over MCP can do, tool by tool. This page is generated from the live tool registry, so if a capability is not listed here the assistant cannot do it yet.

The MCP server exposes six tools of its own — tool_survey, panel, demographics, presentation, payment_confirmation and chat — documented with parameters in the API Reference. The first of those, tool_survey, hands your brief to the surveys agent, which has the internal toolset below. You never call these inner tools by name: you describe what you want and the agent picks them. Names are listed so you can be precise ("run surveycheck", "give me get_crosstab of Q3 by Q7") and so you can tell what is and is not possible.

Legend. ⚠ marks an action the agent will not take on the first ask: it tells you what is about to happen and waits for a confirming message on the same conversation_id before doing it (publishing, deleting, sending email, spending money).

What the surveys agent can do

Plan

Before anything is built.

ToolWhat it does
plan_surveyPresent a structured survey plan for user approval BEFORE building.

Research

Ground the questionnaire in real-world context.

ToolWhat it does
attach_research_sourcesPersist research citations on a survey so they survive outside the chat.
get_research_sourcesRead the research citations stored on a survey.
research_audienceResearch a target audience — demographics, psychographics, behavior patterns, language preferences.
research_competitorsResearch competitors in a given industry or product category.
research_marketDo a quick web research pass on a market or topic.

Templates

Start from one, or save one.

ToolWhat it does
create_survey_from_templateCreate a new survey from a template.
list_templatesList available survey templates.
save_survey_as_templateSave a survey's current design as a reusable template.

Surveys

Create, read, update, lifecycle.

ToolWhat it does
attach_surveyBring an EXISTING team survey into THIS chat as a reference so you can view or work on it here — use when the user points at a survey created in another chat ('pull up my Nike survey', 'work on the BNPL one').
clone_surveyCreate an exact copy of an existing survey as a new DRAFT.
complete_survey ⚠Mark a survey as COMPLETED — permanently closes response collection.
create_surveyCreate a new empty survey in DRAFT status.
delete_survey ⚠Soft-delete a survey.
get_surveyRetrieve a survey with its full document (all blocks, questions, logic, settings).
list_surveysList surveys — by default only the current user's, optionally the whole team's.
pause_survey ⚠Pause an ACTIVE survey — temporarily stops accepting responses.
publish_survey ⚠Publish a survey — transitions it from DRAFT or PAUSED to ACTIVE.
set_redirect_urlsSet the survey's exit redirect URLs — where respondents are sent after finishing, per terminal status (complete / disqualify / terminate / over_quota).
set_url_variablesDeclare entry-link (URL) variables — Decipher's extraVariables.
transfer_survey ⚠Transfer OWNERSHIP of a survey to the current user (you become its owner).
update_surveyUpdate a survey's metadata (name, objective) and/or its full document.

Questions and blocks

Author and restructure the questionnaire.

ToolWhat it does
add_blockAdd a new empty block (section) to a survey.
add_questionAdd a question to a survey.
change_question_typeConvert an existing question to a different type in one atomic operation.
clear_block_rotationRemove block rotation.
create_answer_listCreate a reusable Answer List — an option list defined once and referenced by many questions (agreement scales, brand lists, country lists).
delete_answer_listDelete a reusable Answer List.
delete_block ⚠Delete a block (section) and ALL questions inside it.
duplicate_blockDuplicate a block and all its questions.
get_question_typeEverything about ONE question type: its config keys with defaults and allowed values, the keys that must NOT be used (and what replaced them), what logic can do with its answer, and a complete worked example of the question as stored.
move_blockMove a block (section) to a new position in the survey's block sequence.
move_questionMove ONE question into a block (optionally at a position inside it).
remove_questionRemove a question from a survey.
rename_blockRename a block (section).
rename_questionRename a question's label AND rewrite every reference to it across the survey in one atomic operation: display conditions, skip conditions, skip targets, variable definitions, and option piping (linked_options) are all updated together — nothing breaks.
reorder_questionsReorder all questions in a survey.
rotate_blocksRotate (shuffle) BLOCK ORDER per respondent: the listed blocks form one rotation group and are shuffled per respondent, session-stable.
update_answer_listUpdate a reusable Answer List's name and/or REPLACE its options.
update_blockUpdate a block's config: shuffle the questions inside it per respondent, or loop the block once per selected answer or list item.
update_questionUpdate an existing question's text, options, config, or logic.

Quotas

Targets, least-fill, sheets, progress.

ToolWhat it does
add_quotaAdd a single quota without touching existing ones.
delete_quota ⚠Delete a single quota by ID.
export_quota_reportGenerate a quota status report — quota targets, fill counts, fill percentages, basic analytics.
get_quota_progressGet live quota progress — target vs actual fill for each quota, with fill percentages and remaining counts.
reconcile_quotasRecount all quotas from actual response data and sync Redis.
set_quotas ⚠Replace ALL quotas for a survey.
update_quotaUpdate a single quota's name, target, soft floor, least-fill priority or SHEET.

Variables and keymaps

Entry-link variables and the team's coding keys.

ToolWhat it does
get_variables_keymapGet the team's variable keymap — a dict mapping survey variable labels to custom coding keys used by the team's downstream systems.
set_variables_keymapMerge key-value pairs into the team's variable keymap.

Languages and AI review

Translation and AI-assisted quality.

ToolWhat it does
add_survey_languageMake THIS survey multi-language: add a language as an in-place translation overlay (the survey stays ONE survey — respondents pick their language at the start; every question renders in it).
analyze_open_endsAI analysis of open-ended (text) responses.
optimize_surveyAI quality review of a survey.
set_translationChange ONE translated string on a multi-language survey — the way to act on a client's comment left while reading a translation (the comment carries language).
summarize_resultsAI key takeaways for a survey — structured topic clusters with evidence, suitable for the FE 'AI Insights' panel.
translate_surveyTranslate a survey into another language.

Validation

What would stop a publish.

ToolWhat it does
publish_readinessWhat would stop publish_survey right now, in the gate's own words: blockers (cannot publish), errors and publish-readiness items (need the user's explicit acknowledgement: no QA run on this revision, no quotas, placeholder wording).
surveycheckValidate a survey before publishing.
validate_dslParse a DSL expression and report whether it is syntactically valid.

QA and field checks

QA rounds, path walks, field reports, client review.

ToolWhat it does
get_field_reportThe field report for a live or finished survey — real respondents replayed through the flow proof.
list_qa_findingsThe QA ROUND ledger — only what a recorded QA round or a reviewer filed.
qa_overviewOne call for 'what's the fielding status?' and 'what's open in QA?'.
run_qa_roundStart a QA round (the same as Run QA in the builder's QA tab) on the version being edited and wait for it.
send_review_linkCreate a client review link and email it.
update_qa_findingUpdate one or more QA findings in the ledger: status (open / accepted / fixed / dismissed), assignee, note.
walk_survey_pathWalk one path through the survey exactly as the runner would and explain every decision: which questions each page shows and hides (with the condition and its result), the answers given, variables assigned, the skip that fired and where the path ends.

Versions

Pending vs live versions.

ToolWhat it does
compare_versionsCompare two versions of a survey — shows what questions were added, removed, or modified between them.
get_versionLoad a specific version of a survey document.
list_versionsList all versions of a survey — version number, who changed it, when, and a summary of changes.
push_version ⚠Push a pending version live — respondents will now see this version.
restore_version ⚠Restore a previous version of the survey document.

Edit history and undo

Who changed what, and undo it.

ToolWhat it does
survey_historyRecent EDIT history for a survey — who changed what, when, and from where (chat vs the builder).
undo_survey_change ⚠Undo a previous edit to a survey — including edits made in the builder, not just your own.

Paid sample (Prolific, Dynata, Cint)

Quote, pay, launch, monitor.

ToolWhat it does
get_attribute_optionsDrill into the options of a SINGLE panel attribute.
list_panel_filtersList every targeting filter a panel provider supports — the keys you can put under targeting in panel_quote, plus the canonical values each key accepts.
list_panel_integrationsList every panel-vendor integration on a survey — each Prolific, Dynata or Cint buy with its status.
list_panel_marketsWhich countries a provider fields in, and IN WHICH LANGUAGES.
list_panel_providersList the panel providers available on this platform (currently Prolific, Dynata, and Cint).
panel_pause ⚠Pause an active panel — stops sending new respondents to the survey but doesn't refund or close the vendor project.
panel_quoteGet a price + feasibility quote from a panel provider WITHOUT committing money.
panel_resume ⚠Resume a paused panel — vendor starts sending respondents again.
panel_statusGet LIVE status from the panel provider — completes so far, starts, in-progress, disqualified, over-quota counts.
request_panel_payment ⚠Collect payment for launching a panel buy, then launch it once the user pays.
search_panel_attributesSearch the vendor's FULL attribute catalog for non-standard targeting.

Distribution: email, your own panels, sample sources

Email lists, first-party panels, BYO panel vendors.

ToolWhat it does
add_panel_membersAdd people to an existing panel.
add_recipients_to_email_listAppend new recipients to an existing email list.
attach_panel_to_surveyAttach a panel to a survey, so it becomes part of the project rather than a one-off send.
attach_sample_sourceAttach a library sample source to a survey — COPIES the source's redirects onto the survey's runtime settings and returns the entry link fragment (?fp_src=<id>&rid=) the partner appends to the survey's public URL.
compose_email_templateDraft (or rewrite) a survey-invitation email using AI.
create_custom_panelCreate a reusable custom panel — a saved list of email addresses that can be used across multiple surveys.
create_email_listCreate an email distribution list for a survey.
create_first_party_panelCreate a panel from people the user listed in the conversation.
create_sample_sourceSave a BYO sample partner (external panel company) to the team library: name + per-status exit-redirect URL templates (complete / disqualify / terminate / over_quota / bad_quality) + any extra entry params.
delete_sample_source ⚠Delete a sample source from the team library.
detach_panel_from_surveyRemove a panel from a survey.
detach_sample_source ⚠Remove an attached sample source from a survey.
get_email_list_statsGet delivery and response statistics for an email list — how many sent, opened, started, completed, bounced.
get_panel_send_reportHow a panel send is doing: sent, started, completed, bounced.
list_custom_panelsList all custom panels for the current team.
list_first_party_panelsList the team's OWN panels — contact lists they uploaded themselves, as opposed to synthetic personas or a vendor buy.
list_sample_sourcesList the team's saved sample sources (BYO panel companies: name + per-status exit redirects + entry params).
preview_panel_audienceCount who would actually be emailed, WITHOUT sending anything.
resend_email_list ⚠Resend an email campaign to recipients who haven't completed yet.
send_email_list ⚠Trigger sending of an email list — delivers survey invitations to all recipients in the list.
send_platform_email ⚠Email platform-generated files (surveys, reports, response exports) to up to 10 recipients.
send_survey_to_panel ⚠Email a survey invitation to everyone on a first-party panel.
update_sample_sourceUpdate a library sample source (name / redirects / entry params / notes).

Data quality

Fraud screening and response hygiene.

ToolWhat it does
detect_speedersFind responses that were completed suspiciously fast.
detect_straightlinersFind responses where the respondent gave the SAME answer to every row in grid questions.
generate_test_dataGenerate JUNK fake responses for a survey — randomised answers with NO connection to any persona, panel, or research design.
get_quality_summaryThe fraud and data-quality summary for a survey: how many entrants Research Defender screened, blocked (with the top reasons) or flagged, and how ReDem scored the completes.
set_data_validationTurn the paid data-validation tools on/off for a survey: Research Defender (pre-survey fraud/bot/duplicate screening, billed per check) and ReDem (post-survey response-quality scoring, billed per complete).

Analysis, conjoint, and exports

Frequencies, crosstabs, NPS, conjoint simulator, files.

ToolWhat it does
conjoint_interaction_testTest whether pairs of attributes interact (for example brand × price) rather than acting independently.
conjoint_optimize_productFind the best product to launch: every combination of levels scored against the competitors, ranked by share or by revenue (share × price, when the design has a price attribute).
conjoint_segmentsSegment the respondents by what they want.
conjoint_sensitivitySensitivity analysis: one product with each level of each attribute swapped in turn, against the competitors — how its share moves with price, brand, etc.
conjoint_willingness_to_payWillingness to pay for each level of each non-price attribute, in the price attribute's units: the median premium respondents would pay over the attribute's least-preferred level.
create_conjoint_share_linkCreate a no-login link a client can open the conjoint simulator with — one question, one audience, shares computed on our side, response rows never leave.
exclude_response ⚠Exclude (or re-include) a response from analytics.
explain_conjoint_methodsWhat the conjoint simulator computes and how: the estimation (HB), the fit diagnostics and the convergence policy, the reporting scales, the share rules, the none option, every analysis, the design report, the outputs, and how it compares with Sawtooth.
export_csvExport survey response data as CSV.
export_docxExport the survey INSTRUMENT (questions, options, logic) as a Word document (.docx).
export_pdfExport the survey INSTRUMENT (questions, options, logic) as a PDF document.
export_xlsxExport survey response data as an Excel workbook (.xlsx): a Data sheet with the wide two-row header, a labeled sheet, and a Questions codebook sheet.
generate_csvGenerate a CSV file from agent-composed rows and return a presigned download URL (valid 1 hour).
generate_docxGenerate a Word document (.docx) from agent-composed content and return a presigned download URL (valid 1 hour).
generate_pdfGenerate a PDF document from agent-composed content and return a presigned download URL (valid 1 hour).
generate_xlsxGenerate an Excel workbook (.xlsx) from agent-composed content and return a presigned download URL (valid 1 hour).
get_all_frequenciesGet answer distributions for ALL questions in the survey at once.
get_analyticsGet a summary of survey response data: total responses, completion rate, median time, status breakdown, daily timeline (last 30 days), and country breakdown.
get_crosstabTwo-dimensional cross-tabulation: rows = one question's answers, columns = another question's answers, cells = count.
get_frequenciesGet the answer distribution for a single question — how many people picked each option, with counts and percentages.
get_npsCalculate the Net Promoter Score for an NPS question.
get_responseGet the full detail of a single response — every answer, timestamps, geo, panel info and the order in which blocks, questions and options were shown to that respondent.
list_conjoint_scenariosThe scenarios saved on a conjoint simulator card, with the shares they scored when saved.
list_responsesList individual responses for a survey — paginated, filterable by status.
save_conjoint_scenarioSave the CONJOINT SIMULATOR scenario just discussed (products, rule, none-of-these set-up, filters) under a name — 'save this as …', 'remember this set-up', 'keep this scenario' after a simulation.
show_conjoint_simulator_cardOpen the conjoint simulator for a conjoint question, as a card in the chat.
show_results_cardPut a survey's live results dashboard into the chat as a card, scoped to ONE audience bucket.
simulate_conjoint_marketRun the conjoint simulator: give it products as attribute -> level text and get each one's share of preference (optionally with a 95% interval), under a share rule, with or without a none-of-these alternative, for all respondents or a filtered segment.
stat_testRun a chi-square test of independence on a cross-tabulation.

Comments

Review threads on a survey.

ToolWhat it does
add_commentAdd a comment to a survey — optionally on a specific question.
list_commentsList all comments on a survey — shows text, who posted, which question (if any), and whether it's resolved.
resolve_commentMark a comment as resolved — indicates the feedback has been addressed.

Programming preferences

How this team likes surveys built.

ToolWhat it does
get_programming_preferencesRead how this team programs surveys — the conventions they have agreed and accepted, including the markers their documents use and the exact strings those become.
remember_programming_preferenceFile something the user told you about how they want surveys programmed, so it is applied next time.

That is 146 tools. The agent chains them itself: "build a 12-question NPS survey with an age quota, translate it to Spanish, check it, and publish" is one brief, not five calls.

What the other MCP tools do

ToolCapability
panelBuild a synthetic panel of AI personas (up to 200), run a fresh panel through a survey in one call, or recruit real paid respondents through the surveys agent's quote → confirm → checkout flow.
demographicsPopulation, age, race, gender, education and income distributions for any place, with AI insights.
presentationA native PowerPoint deck with editable charts from any data or analysis.
payment_confirmationSettle an x402 USDC payment for a panel launch. Autonomous wallet agents only; humans pay through a Stripe Checkout link.
chatThe full orchestrator. Use it for multi-domain plans and for the capabilities that have no direct tool: ad and market experiments on Google, Meta and LinkedIn Ads (build variants, test them on synthetic personas, launch, analyze), research scoping and Statement of Work drafting, asking an existing synthetic panel a question, listing or calibrating the team's panels, the team file library, and maps of places or regions.

What is deliberately not available over MCP

Some things need a browser, a file upload, or are internal to the agent. Knowing this up front saves a round trip.

Not over MCPDo this instead
Uploading a spreadsheet of contacts or a questionnaire document (DOCX import)Console → My Panels for contacts; Console → Import for a questionnaire. The agent cannot read files from a chat.
Interactive cards: plan review, approval card, results dashboard, conjoint simulator, Stripe payment panelThe agent describes the same content in text. The cards render in the web app, and every MCP conversation is visible at app.flashpoint.ai. Paid recruitment returns a Stripe Checkout link to open in a browser.
Qualtrics, Confirmit and Forsta export formatsConsole → Export, or the surveys REST API. CSV / XLSX / DOCX / PDF are available through the agent.
Running an existing synthetic panel by id through panelpanel(action="create_and_run"), or ask chat to run the panel by id.
Two tools the agent uses internally (redirect_to_orchestrator, report_unsupported_request)Nothing to do: the first hands non-survey requests to the orchestrator, the second files a feature request with engineering when something genuinely cannot be done.

Limits that apply over MCP

LimitValue
Email invitations + reminders per team per week100, resets Monday 00:00 UTC; raised per team on request to help@flashpoint.ai
Email sends per conversation per survey50
Synthetic panel size200 personas
tool_survey call durationup to 600 s; long jobs stream progress notifications
Wallet agentsper-call USDC pricing, see API Reference › Pricing

See Email distribution › Sending limits for how the email budget behaves and what a refused send looks like.