Origin by Dialectica · MCP connector

Origin MCP Documentation

Works with Claude & any MCP clientOkta OAuth 2.1 — no API keys
Getting started

Connect in minutes

The connector requires a Dialectica Origin account. Access is enterprise / institutional only — not yet on Origin? Contact your Dialectica account team.

1

Add the connector

In Claude: Settings > Connectors. Find Origin by Dialectica in the directory, or add it as a custom connector with mcp.dialecticanet.com/mcp. Any MCP client that supports remote servers can use the same URL.

1

Add the connector

In Claude: Settings > Connectors. Find Origin by Dialectica in the directory, or add it as a custom connector with mcp.dialecticanet.com/mcp. Any MCP client that supports remote servers can use the same URL.

2

Sign in with your Origin account

Your AI platform opens the Origin sign-in (hosted on Auth0). The OAuth 2.0 handshake does the rest — nothing to copy, nothing to rotate. Signing in is also what unlocks search across the transcripts from all your past Dialectica projects.

Team and Enterprise admins: the connector can be enabled for your organisation from the Claude admin settings, so users only need to sign in with their Origin account.

Example prompts

Ask like an analyst

Four prompts that show the range — from one-shot synthesis to launching a full expert-calls project without leaving the thread.

01COMPANY PROFILE

“Give me AppsFlyer’s profile — financials, products, business model and what customers say.”

Resolves to AppsFlyer (appsflyer.com, PE-backed, Business Intelligence Software) and returns the structured profile: introduction, product lines, latest-year financials, pricing and go-to-market, customer insights, and former-executive interviews — plus name lists of customers and competitors with a pointer to the detail tools.

origin_get_companyEvery sourced field carries a link to its Origin source · typically a few seconds
02Competitive landscape

“Who competes with AppsFlyer, and what do experts say about the positioning?”

An evidence-ranked competitor table — each row with an Origin profile link, industry, market and headcount or revenue when known — followed by expert quotes that name the competitor and link to the source.

origin_find_competitorsRanked by evidence tier · capped at 50 rows with shown / total
03SCREEN & RANK

“Find VC-backed marketing-software companies in the US with more than 500 employees.”

The criteria echoed back (industry, ownership, location, headcount floor) and a ranked list of matching companies — name, profile link, one-line introduction, revenue, EBITDA and FTEs where known — with total_matched and shown counts.

origin_search_companiesStructured filters, a free-text theme, or both
04Find experts

“Which former AppsFlyer executives left in the last three years?”

A roster of experts in Dialectica’s network tied to AppsFlyer — name, role, seniority code, tenure dates, perspective (former / current / competitor) and a LinkedIn link where known — filtered to former employees whose role ended within three years.

origin_find_expertNo source citations — experts are people, not data points · total_eligible and truncated tell you when the list was cut
05SYNTHESIS

“What’s driving AppsFlyer’s growth — and where is it most vulnerable?”

The criteria echoed back (industry, ownership, location, headcount floor) and a ranked list of matching companies — name, profile link, one-line introduction, revenue, EBITDA and FTEs where known — with total_matched and shown counts.

ask_originMarkdown · one query in flight per user at a time
06YOUR PROJECT TRANSCRIPTS

“What did the experts on my project say about switching costs in mobile attribution?”

Cited excerpts drawn from the transcripts of all your past Dialectica projects — plus a shared pool — each with a stable [short_id], the company discussed, and a scope marker: your_project or shared.

query_private_transcriptRequires Origin sign-in with project access · truncated=true means a size-limited subset was returned

Capabilities

What the connector can do

Seven read-only tools. Six direct lookups return structured JSON in seconds; one research agent synthesises across the whole corpus. Expand any row for its arguments, response and limits. Nothing here writes to Origin, places an order, or spends credits.

Research & synthesis

1 tool
ask_origin Ask an open-ended company or market question. Runs Origin’s research agent and returns a cited Markdown answer with sources and suggested follow-ups. Best for multi-company, thematic or comparative questions; slow (typically 60–180 s). Read
Arguments
ArgumentTypeRequiredDefaultDescription
querystringyesThe company or market question to research.
advanced_searchbooleannofalseWhen true, also searches proprietary expert-call transcripts. Requires signed-in access; returns an explanatory error if the caller has no access token.
Returns

Markdown string — answer prose with inline [n] markers, then ## Sources ([n] title — url), ## Suggested follow-ups, and ## Report (download link) when a PowerPoint report was generated. Each of the three sections is omitted when empty.

Notes
  • Queries above 1,000 tokens are rejected with an explanatory message instead of running the agent.
  • One ask_origin call in flight per user; a concurrent call returns “An Origin query is already in progress for this caller.”
  • Requests time out at 900 s with “The request took too long to complete.”
  • For a single named company, the direct lookups below are faster and more precise.

Company intelligence — direct lookups

5 tools
origin_get_company Profile of one named company: overview, financials, products, markets, business model, customer insights, interview evidence and recent news. Each sourced field links to Origin. Read
Arguments
ArgumentTypeRequiredDefaultDescription
company_namestringyesCompany name to resolve.
Returns

(CompanyProfilePayload, additive-only): the standard found / query / resolved fields plus name, headquarters, year_founded, domain, ownership_type, introduction, products[], industries[], markets[], financials[] (by year: revenue, revenue growth, EBITDA, EBITDA margin, revenue per FTE, total FTEs — each with sources), business_model (pricing, go-to-market, contract length, average contract value, sales cycle, growth review, revenue breakdown), customer_insights[], interviews (count, date range, seniority, profile link, excerpt chunks), customers and competitors (name lists capped at 50 with a total and a detail_tool pointer), investors, acquirers, recent_news[].

Notes
  • customers / competitors here are name pointers only — call origin_find_customers / origin_find_competitors for the full, evidence-ranked, source-cited rows.
origin_find_competitors Evidence-ranked competitors of a named company, with profile links, financials when known, and supporting expert quotes. Read
Arguments
ArgumentTypeRequiredDefaultDescription
company_namestringyesCompany name to resolve.
Returns

(CompetitorsPayload, additive-only): the standard found / query / resolved fields plus competitors[] (name, id, profile_url, domain, year_founded, ownership_type, industries, markets, introduction, financials {year, currency, revenue, revenue_growth, ebitda, ebitda_margin, total_ftes}, sources[]), shown, total, truncated, full_list_url, and quotes[] (competitor_name, text, source).

Notes
  • Rows are ranked by evidence tier — private transcripts / interviews and expert quotes first, financials as tiebreaker — and capped at 50. When cut, shown, total, truncated and full_list_url tell you where to get the rest.
origin_find_customers Customers of a named company, with profile links, customer insights and supporting quotes. Read
Arguments
ArgumentTypeRequiredDefaultDescription
company_namestringyesCompany name to resolve.
Returns

(CustomersPayload, additive-only): same top-level fields and row shape as competitors — customers[], shown, total, truncated, full_list_url, quotes[] — plus insights[] (title, value, scale, goal_low, sources).

Notes
  • Same evidence ranking and 50-row cap as origin_find_competitors.
origin_find_experts Experts in Dialectica’s network tied to a named company — former, current or competitor perspectives — filterable by seniority and recency. LinkedIn link where known. Read
Arguments
ArgumentTypeRequiredDefaultDescription
company_namestringyesCompany name to resolve (full legal name, abbreviation or common shorthand).
perspectivearray[string]noformer, competitor, current. Pair cut_off_years with former and/or competitor, not current.
seniority_levelsarray[string]noSeniority codes c, c1, c2, c3.
cut_off_yearsintegerno4 (API default)For former / competitor only: include experts whose role ended within this many years.
filter_company_idsarray[string]noRelated-employer company UUIDs to filter by.
Returns

(ExpertsPayload, additive-only): the standard found / query / resolved fields plus experts[] (id, name, role, seniority, company, start_date, end_date, current, perspective, location, linkedin_url), shown, total_eligible, total_fetched, truncated.

Notes
  • No sources array — experts are people, not cited claims. LinkedIn URLs let users verify the person.
origin_search_companies Ranked list of companies matching structured filters (industry, ownership, location, revenue / EBITDA, headcount) and/or a free-text thematic query. Read
Arguments
ArgumentTypeRequiredDefaultDescription
industryarray[string]noIndustry labels as stated; the server normalises against the Origin taxonomy and suggests alternatives for unrecognised values.
ownershiparray[string]noAcquired, Founder-owned, PE-backed, Public, VC-backed.
locationarray[string]noRegion or country names (e.g. USA, UK, LATAM); normalised, with suggestions for unrecognised values.
financialsobjectnoRevenue and EBITDA ranges in millions of US dollars: { revenue: {min, max}, ebitda: {min, max} }.
ftesobjectnoHeadcount (full-time equivalent) range: { min, max }.
semantic_querystringnoFree-text thematic query for hybrid semantic search.
Returns

(SearchResultsPayload, additive-only — no query / resolved, since this is not a name lookup): found, message, error, criteria (echo of the filters applied), results[] (id, name, profile_url, domain, year_founded, ownership_type, introduction, products[], industries[], geographic_markets[], revenue, ebitda, ftes), shown, total_matched, truncated.

Notes
  • At least one filter or a non-empty semantic_query is required; otherwise found=false with a message asking for criteria.

Your Dialectica project transcripts

1 tool
query_private_transcript Ask a question across the expert-call transcripts from all your past Dialectica projects — plus a shared pool — and get cited excerpts back. Requires Origin sign-in with project access. Read
Arguments
ArgumentTypeRequiredDefaultDescription
querystringyesQuestion to search entitled transcripts for.
Returns

(PrivateTranscriptPayload, additive-only — no resolved, since there is no company-name step): found, query, excerpts[] (short_id, excerpt, company_name, scope = your_project | shared, timestamp), shown, total, truncated, message, error.

Notes
  • Covers the expert-call transcripts from all your past Dialectica projects, plus a shared pool that is always included (scope shared).
  • Requires Origin sign-in with project access; refused otherwise, with a plain explanation.
  • Cites by [short_id], not URL — transcripts have no page of their own. Full transcripts (audio and PDF) are available in Origin; this tool is fast Q&A over excerpts, not export.
  • A different index from the instant transcripts surfaced inline by origin_get_company.

Read-only means the connector only retrieves data. Claude can call these tools without a per-call confirmation prompt, and none of them can modify your Origin account, launch a project, request a report or spend credits. Requests, snapshots and expert calls are placed in Origin itself.

How it works

Grounded by design

A few behaviours hold across all seventeen tools. Knowing them makes the answers easier to trust, and easier to debug.

Company resolution

Every company tool runs the name through search_companies first, so typos and half-remembered names still land on the right Origin profile. When a name matches more than one company, you get the candidates back and pick, rather than an error.

Citations travel with the answer

ask_origin_ai synthesises across the company, market and expert corpus and cites as it goes. get_company_profile returns the full company page as an interactive card, Sources tab included, so you can follow any claim back to where it came from.

Three levels of permission

Every tool carries a badge. Read retrieves and changes nothing. Write creates or updates something in Origin, such as a project brief or a snapshot request. App-only covers the two purchase tools, which the model can see but can never call.

Nothing is bought without you

purchase_instant_snapshot and purchase_transcript run only when you press the button inside the card, with the cost already on screen. No phrasing of a request gets Claude past that button.

Research keeps its place

Work lives in threads rather than in one disposable answer. create_conversation opens one, list_conversations shows what you have been working on, and get_conversation reopens a thread with its messages, sources and follow-ups intact.

What actually gets sent

Each call sends its own arguments, a question or a company name and filters, along with your sign-in token. The connector does not read your conversation history, your files or your memory. How Dialectica handles what it receives is set out in the Privacy Policy.

Troubleshooting

When something looks off

Most issues are one of these. Each entry gives the symptom, the cause and the fix.

Several companies come back instead of the one you meant
Cause

search_companies matched the name to more than one Origin profile.

What to do

Re-ask with something that separates them: the domain, the HQ country, or the sector. Working through every candidate in turn is slower than naming the right one.

A company resolves but a lens comes back empty
Cause

Origin holds the company but has nothing on file for that particular view, for example no named customers.

What to do

Try a different lens, such as get_company_information or get_competitor_information, or ask ask_origin_ai for a synthesised view across the corpus.

Claude describes a purchase but nothing is bought
Cause

Working as intended. Both purchase tools are app-only, so the model can surface the option but cannot act on it.

What to do

Press the button inside the card yourself. The cost appears before you commit.

A project brief was written but never launched
Cause

start_project_brief and update_project_brief only build the draft. Nothing reaches Dialectica until you confirm.

What to do

Finish the follow-up questions, then confirm the submit_project_brief step. Until you do, the brief sits as a draft.

Validation or unexpected-properties errors after an Origin update
Cause

MCP clients cache tool definitions when they connect, so an older copy of the schema is still in play.

What to do

Fully restart the client. Quit and reopen it, because reconnecting on its own does not refetch the schemas.

A list looks shorter than it should be
Cause

Long lists are capped so a response stays readable. [CONFIRM THE ROW CAPS]

What to do

Open the company in Origin for the complete list, or narrow the question so the rows you care about rank higher.

filter_companies returns nothing useful
Cause

No criteria were supplied, or the combination of industry, geography, ownership and financial ranges is too tight to match anything.

What to do

Give at least one filter, then loosen a single range at a time, usually revenue or headcount, until results appear.

Sign-in loop, or connector unavailable
Cause

The session expired, or the account is not provisioned for Origin.

What to do

Sign out and back in from the connector settings. If it keeps happening, write to origin_support@dialecticanet.com.

Support

Get help, report an issue

Where to go for product questions and security concerns, and the policies that govern the connector.

Contact

Product support

Questions about setup, access or results: origin_support@dialecticanet.com. Include your Origin account email, the tool name and the question you sent.

Contact

Security

Report a vulnerability or security concern: security@dialecticanet.com. Please keep client data out of the first message.

Policy

Privacy

How Dialectica collects, uses and retains data: Privacy Policy.

Policy

Terms

Use of the connector is subject to your Origin agreement and Dialectica’s Terms & Conditions: Terms.

Origin by Dialectica is operated by Dialectica. The connector calls Dialectica’s own APIs at mcp.dialecticanet.com and does not proxy third-party services. Server: https://mcp.dialecticanet.com/mcp · Authentication: OAuth 2.0 (Origin sign-in) · Tools: 17 across four groups, covering read, write and app-only purchases.

Put a research analyst in every thread

Connect Origin and your next question comes back grounded, cited, and one message away from a full Dialectica project.

Connect Origin