Use casecompany research agent

Company research agent with official source evidence

Give an AI agent a repeatable company-research workflow built on official profiles, filings, documents and explicit limitations.

A reliable company research agent separates entity resolution, source retrieval, document review and interpretation. dossaro supplies the source-backed MCP operations and evidence references; the calling agent defines the assignment, composes the findings and keeps missing or restricted evidence visible.

Updated July 30, 2026

Run this exact task

Install dossaro, then continue the research

The setup guide lets a capable agent add the remote MCP, start OAuth with the user and return to this task.

Install-and-run prompt
Review the dossaro MCP setup guide at https://dossaro.com/agent-setup/prompt.md, configure dossaro when possible, complete OAuth with the user, then perform this task: Research this company or target list by resolving every legal entity, retrieving the strongest supported official profile, relationship, filing, document, insolvency and regulatory evidence, citing each source and date, and returning ambiguity and access limitations without inventing ownership or diligence conclusions.

Direct answer

A company research agent is useful when it executes a defined evidence workflow rather than returning an unreferenced company summary. The agent should first resolve the legal entity, then select the official source operations that match the question, preserve returned references and finally explain what the evidence does and does not establish.

dossaro provides those source-backed primitives through its Research MCP. It does not autonomously decide the correct acquisition target, beneficial owner, investment conclusion or legal interpretation.

Define the assignment

State the target company or list, jurisdiction, relevant date, evidence questions and acceptable source limits. Distinguish deterministic retrieval from the analysis the calling agent must perform.

Match names to source-native identifiers and retain ambiguous candidates. A brand, domain or trading name is not enough to join records across jurisdictions.

Route by evidence question

  • use company profiles for legal identity and status;
  • use relationship records only for the roles the source actually exposes;
  • use company-register documents for acts, accounts, filings and source artifacts;
  • use company insolvency notices for dated official proceedings evidence;
  • use regulatory sources such as BaFin for supervised-firm or fund evidence;
  • use publication sources such as BORME for dated registry events.

Compose outside the source call

The agent should compare evidence, state dates, separate direct observations from inferences and return a manual-review queue. Missing public evidence must remain a limitation rather than becoming a negative finding.

Inputs and outputs

Recommended inputs:

  • legal name, known identifier and expected jurisdiction;
  • research question and relevant date;
  • required profile, relationship, filing, document, insolvency or regulatory evidence;
  • whether historical changes matter;
  • output schema and manual-review rules.
OutputPurpose
Resolved entityShows which legal company the agent researched
Source-operation logRecords which official surfaces were queried
Evidence referencesKeeps returned profiles, notices and documents reviewable
Dated findingsPrevents stale evidence from being described as current
LimitationsDistinguishes blocked, paid, restricted, unavailable and no-result states
Review queueIdentifies ambiguity or interpretation requiring a person

For French work, the RNE company and document workflow illustrates how entity resolution and account-backed documents remain separate steps.

Jurisdiction differences

  • Germany can combine company-register documents, annual-report publications, insolvency notices and BaFin regulatory evidence.
  • France separates Recherche Entreprises profiles, RNE documents and BODACC announcements.
  • Spain and Portugal expose public mercantile publications without turning those notices into complete free company files.
  • Latvia can expose public member or beneficiary rows while withholding document binaries.
  • Czech ARES can expose linked-register roles and registered shareholder or partner entries without beneficial-owner conclusions.
  • Source access, identifiers, public document types and relationship meanings vary by country.

The agent must therefore route by jurisdiction and evidence surface. A uniform output table is useful only when each row retains its source-specific meaning and limitations.

Review limitations

  • dossaro does not make final ownership, UBO, cap-table, legal or investment determinations.
  • Officer, director, representative, member, shareholder and beneficiary fields are not interchangeable.
  • Public registers can lag private transactions and later filings.
  • Some documents are paid, restricted, account-bound, blocked or metadata-only.
  • Parsed documents can lose visual context and require original-file review.
  • A source error or empty result must not be treated as proof that evidence does not exist.
  • Batch research requires row-level citations and ambiguity handling.
  • The calling agent and user remain responsible for final analysis and decisions.

Run this exact task

Install dossaro, then continue the research

The setup guide lets a capable agent add the remote MCP, start OAuth with the user and return to this task.

Install-and-run prompt
Review the dossaro MCP setup guide at https://dossaro.com/agent-setup/prompt.md, configure dossaro when possible, complete OAuth with the user, then perform this task: Research this company or target list by resolving every legal entity, retrieving the strongest supported official profile, relationship, filing, document, insolvency and regulatory evidence, citing each source and date, and returning ambiguity and access limitations without inventing ownership or diligence conclusions.

Capability graph