BIA Edge

Fetch the next ranked tranche of an earlier search. Use this when a search response reports meta.has_more with a meta.next_cursor token: it continues the same ranked result list where the previous window ended. Green searches read a retained bounded ordering and refresh treatment warnings (caution or null); a missing warning does not establish good law. Lost windows require a fresh search. It never changes the query — to search differently, call the search tool again. Continuation depth is capped (rank offset 100) because deeper ranked tails are low-relevance; an expired or edited cursor returns a coded error and the original search must be re-run for a fresh cursor.

get_more_results

Rank immigration cases by citation count within the Passim AI corpus. This is a corpus citation-count ranking, not an editorial judgment that a case is legally controlling, important, or still good law. Args: source: Filter by source: 'bia', 'aao', 'ag', 'circuit-courts' limit: Max results (default 25)

get_most_cited

Fetch exact evidence spans by IDs returned from search_passages. When a returned span's case has propositions, the response also carries a top-level propositions list (one block per document with the propositions, a propositions_note, and counts, bounded by propositions_limit) so the span can be read against the case's propositions. Propositions are never validity signals; use check_validity to review identified treatment and source evidence. Absence of a warning does not establish good law. Args: span_ids: Span IDs such as case:123:corpus_block:789, case:123:issue_unit:456 or historical case:123:block:789. include_propositions: Attach per-document propositions blocks (default true). propositions_limit: Maximum propositions per document; clamped to 1-100 (default 25).

get_passages

Find state opinions citing a statutory provision and exact subsection. Returns source-bound pinpoint passages and official/document links. Links identify citation mentions; interpretive significance remains unreviewed. State authority is not immigration precedent and has no ImmCite treatment. First use lookup_criminal_code/get_criminal_code_provision with offense_date or conviction_date for the applicable stored statute text and coverage limits. Opinions do not automatically apply the statutory version at decision date.

get_provision_case_law

Get full text for a reference-section record. Args: reference_id: The reference section ID (from search_references results) max_chars: Max characters of full text to return (default 15000) section: Optional heading/subsection title to retrieve from a long reference text_offset: Character offset into the returned view (the section, or the selected subsection) at which the text starts; continue a truncated response with its next_text_offset

get_reference

Resolve a known USCIS, FAM, form-instruction, or practice-reference citation. Use this before search_references when the citation or source label is known. Use find_in_reference for pinpoint snippets or fetch for larger section text after a reference is resolved.

get_reference_by_citation

Check recent BIA/AG decisions, CFR and Federal Register changes, and USCIS policy updates whenever timing may affect the answer. Get recent regulatory and policy changes in immigration law. Includes new BIA/AG decisions, CFR amendments, policy manual updates, and Federal Register notices. meta.window_start reports the earliest timestamp the window reached, so an empty result says how far back it looked. topic accepts a known topic name (the litigation-tracker topics plus doctrinal topics such as unlawful presence, adjustment of status, waivers, criminal grounds, cancellation of removal, motions to reopen, in absentia, voluntary departure, fraud, credibility, representation, and national interest waiver) or free-text terms that must all appear as whole words; a statute section such as 212(a)(9) works as a term, and generic agency words are ignored with a warning naming the terms used. Rows that cannot match the topic are excluded in the database before the candidate window is filled, so unrelated notices cannot hide older matches. update_type and source filter in the database; include_* flags admit material excluded by default. Continue with meta.next_cursor.

get_regulatory_updates

Get the Passim AI legal-research playbook to follow for serious research. Fetch and follow deep_research_method before filing-facing work: legal memos, briefs, authority packets, RFE/NOID support, crimmigration screens, or any research where a wrong answer creates filing or malpractice risk. Fetch unpublished_routing before searching unpublished BIA/AAO decisions. Fetch authority_taxonomy when classifying binding vs. non-precedent authority. Fetch research_method for the overall method: request-mode classification, operating principles, research floors, validity-check discipline, memo structure, and citation rules. Fetch tool_routing for the complete tool-by-tool routing map. Returns the requested section as markdown. The playbook is methodology, not law: it tells you how to research and present authority, and the direct source tools remain the only source of legal content. Follow it yourself; the server does not synthesize final legal work product. Args: section: Which playbook section to return (default research_method).

get_research_playbook

Get the full text of a specific statutory or regulatory provision. Args: provision_id: The provision ID (from search_statutes results)

get_statute

Resolve governing INA, USC, or CFR text from the live corpus before stating the rule. Resolve a known INA, USC immigration, or CFR citation. Use this before search_statutes when the citation is known. Use lookup_criminal_code(jurisdiction='federal') for Title 18 and other federal criminal-code provisions instead of treating a statute search miss as absence. Uncodified statutory notes also resolve by their enacting public-law citation or a verified act name (e.g. "USA PATRIOT Act § 411(c)"); such matches carry metadata.alias_resolution.

get_statute_by_citation

Return Visa Bulletin priority dates loaded in Passim AI, or check one priority date. Without priority_date, returns one bulletin-table record (omit month/year for the latest loaded bulletin, or provide both for a historical month): the record text is the full markdown table, and metadata.columns/rows carry the same data with final-action and dates-for-filing cells as ISO dates or Current / Unavailable / Not Published; metadata.uscis_chart_designations maps categories to the USCIS-accepted chart. With priority_date plus category and chargeability, answers whether that date is available under the selected chart. Use this for stored priority-date research, not case-law research; call get_corpus_info or verify official Department of State sources when the latest bulletin timing matters.

get_visa_bulletin

List the organized district-court decision families and live corpus counts. Use this before district-court MCP searches when you need to choose a practitioner-facing corpus slice. The listed decisions are nonprecedential federal district court decisions and are persuasive only.

list_district_court_decision_families

Search state or federal criminal-code provisions for crimmigration research. Use this before categorical-approach analysis to find the statute of conviction by citation, section number, title, or text. Results include official-source status, parser/ingestion status, and the text for the best match when available. State coverage is penal-title-only for some states, so a miss is not proof a state offense does not exist; the empty response lists the official source records Passim AI holds. search_statutes is the immigration statute/regulation lane and is not proof of criminal-code absence. Use get_criminal_code_provision after lookup when you need one provision's full text and metadata. Dates select stored historical text. Historical-only sections also accept exact archive citations, such as 'MS § 97-1-1' or 'Me. Rev. Stat. tit. 17, § 1', with a date; edition-year coverage is not proof of effective dates. Historical keyword search is not covered. Repeat the same dates and citation when following next_offset.

lookup_criminal_code

Lookup current USCIS/EOIR form-page and form/instruction PDF records. EOIR-numbered forms (EOIR-26, EOIR-28, EOIR-33/IC, EOIR-42B, ...) route to the eoir_forms immigration-court/BIA catalog; all other numbers route to the uscis_forms catalog. Args: form_number: USCIS or EOIR form number such as I-751, I-485, N-400, EOIR-28, or EOIR-42B. max_chars: Maximum source snippet characters per returned source. Defaults compact; use returned reference_id values with get_reference or find_in_reference for full source text.

lookup_uscis_form

Query EOIR statistics rollups as descriptive analytics data. This read-only primitive returns facts, denominators, filters, and caveats. It intentionally does not produce legal strategy, judge predictions, or recommendations.

query_eoir_stats

Review identified treatment of central or dispositive cases in the live ImmCite records before relying on them. ImmCite statuses (each record repeats its own as status_meaning): caution = adverse or limiting treatment found; negative = negative treatment found; overruled = overruled; vacated = vacated or withdrawn; reinstated = earlier adverse treatment undone; unchecked = no stored status, not evidence either way; treatment_available = history exists, not clearance; unchecked_no_treatment_data = eligible, no treatment row; not_applicable = outside the precedential ImmCite corpus. Search shows caution for adverse treatment, otherwise no label; no warning is not good law. Treatment filters: all, positive, negative, distinguishing, overruling; entries keep their stored relationship_type, and status_evidence marks the status's own evidence case. Review identified ImmCite treatment and source evidence for a resolved case document_id. It does not certify that a case is valid: read metadata.status_reason and the entries in metadata.treatments (adverse first; the published status's own evidence case is pinned first with status_evidence true) before relying on an authority. metadata.treatment_available_count and treatment_omitted_count report the window; meta.next_call is the directly callable continuation. metadata.freshness dates the answer: status_computed_at, latest_citing_date, citation_index (citing decisions no treatment entry covers, newest five with links), corpus reach for the subject's source, and newer_citing_call (get_citations for anything decided since); none of these proves nothing newer exists outside the corpus. A warning names unreviewed precedential citing decisions; review them before relying on the status. document_ids (up to 10) returns compact summaries; follow each metadata.detail before relying on one. proposition_key with document_id runs an exact proposition-keyed query.

check_validity

Generic structured search for ChatGPT/Codex MCP clients. Case results expose validity_status="caution" for identified adverse treatment, otherwise null. Absence of a warning does not establish good law. Use check_validity for precise treatment, affected issues, and source evidence. Detail-only coverage states include unchecked, unchecked_no_treatment_data, and not_applicable. Reference searches examine selected sources. Use get_corpus_info(section='sources') for coverage and search_references with source_code to target a specific source. Args: query: Search text. corpus: all, cases, references, statutes, unpublished, categories, or national_library. limit: Maximum results per requested corpus (1-25).

search

Investigate the live immigration case-law corpus for controlling, adverse, limiting, and recent authority. For state opinions, use source='state-criminal-courts' and a required state; names, citations, and keywords search PostgreSQL. For opinions citing a criminal statute, use get_provision_case_law with the state and provision. Search case law held in the Passim AI corpus. Default library_scope="primary" searches the immigration-focused corpus. When primary results do not answer the question, deliberately broaden with library_scope="national_library" for nationwide keyword fallback. National results are grouped by case, retain separate opinion IDs, and have unassessed legal treatment. Fetch their library-opinion IDs with fetch; do not pass those IDs to primary-corpus validity tools. Known citations use get_case_by_citation. A national query that reads like a caption (Miranda v. Arizona, In re Gault) is matched against case titles first, most-cited first (metadata.match_type is case_name); the rest is keyword search of opinion text, where every word must appear (OR, NOT, quoted phrases and parentheses are accepted). Results show caution for identified adverse treatment, superseded_in_same_matter when the only reviewed treatment is the same matter's later final decision (an Attorney General referral order), and otherwise no validity label. Absence of a warning does not establish good law. Use check_validity for precise treatment and source evidence. Use this for names, concepts, or keywords when there is no known citation. Use get_case_by_citation first for a known citation, then bridge central local cases to document_id for check_validity or corpus source-text tools.

search_cases

Search asylum country-condition reports. Searches State Department Country Reports on Human Rights Practices, International Religious Freedom Reports, and Trafficking in Persons reports after those sources are loaded into the reference corpus. Args: country: Optional country name, e.g. "Afghanistan" or "Mexico" query: Optional issue terms, e.g. "political opinion", "FGM", "police torture" report_type: Optional type: hrr/human_rights, irf/religious_freedom, tip/trafficking, or all year_from: Optional earliest report year to include year_to: Optional latest report year to include. Omit to search through the newest loaded report year. limit: Max results

search_country_conditions

Search nonprecedential federal district court immigration decisions. Use this for written federal district judge opinions and orders, including mandamus / agency-delay, immigration habeas, naturalization, USCIS benefits APA review, consular/visa, immigration FOIA, and other civil immigration decisions. The 'mandamus' category is a posture-based lane across subject families. These decisions are persuasive only and are kept out of ordinary precedential search unless district-court material is explicitly requested. Args: query: Search text - issue, agency, party name, docket number, or keywords decision_category: Optional corpus slice, such as 'mandamus', 'habeas', 'naturalization', 'benefits', 'consular', 'foia', 'u_visa', or 'other' court: Optional district court code, such as 'txsd', 'txnd', 'cacd', 'dcd', or 'nysd' date_from: Start date (YYYY-MM-DD) date_to: End date (YYYY-MM-DD) limit: Max results (1-50, default 20)

search_district_court_decisions

Search EOIR immigration judges and return descriptive rollup statistics. This is an analytics primitive. It surfaces denominators, filters, source hash, and caveats; it does not rank judges for case strategy or make judge predictions. When court_code is supplied, returned list metrics are scoped to detail rollups at that court; use get_eoir_profile for career-wide context.

search_eoir_judges

Search materialized source-text spans and issue units. Returns compact evidence records with stable span IDs, offsets, source text, authority metadata, and match reasons. Use this before fetching full cases when the agent needs pinpoint support for a proposition. A statutory citation in the query (237(a)(2)(A)(iii), section 240A) must appear in every returned span; generic words such as under, section, and act are ignored. AAO non-precedent paragraph spans are retrieval-only persuasive support and are returned only when authority_scope='all' is selected; the default precedential scope keeps them out of binding-law searches. When the stored text carries page markers (BIA and Supreme Court page numbers, reporter star pagination, slip-opinion page headers), each span also carries a ``page`` label such as "26 I&N Dec. 551", "546 U.S. 245" or "slip op. at 3", with its basis and confidence; labels at the edges of a text are inferred, and spans without markers carry no page. Args: query: Legal terms, factual phrase, statutory phrase, or quoted language. source: Optional source filter such as bia, ag, aao, circuit-courts, scotus. document_id: Optional document ID to search within one authority. authority_scope: all, precedential, or binding. Default excludes known nonprecedential material. granularity: passage for source-text paragraphs, issue for accepted proposition evidence, or all. limit: Maximum spans to return, up to 50. expected_page: Pin-cite page to verify against the inferred page labels; adds pin_check.

search_passages

Search USCIS, FAM, practice manuals, UNHCR Handbook/guidelines, and asylum officer training. Use this for policy or guidance issues when the exact citation is unknown. Use get_reference_by_citation for known references and find_in_reference or fetch for pinpoint source text after a reference section is identified. Unfiltered searches examine selected sources, not the entire reference corpus. Use get_corpus_info(section='sources') to check coverage before claiming a source is absent.

search_references

Investigate governing INA, USC, and CFR text when the exact citation is unknown. Search INA, USC immigration, and CFR statute/regulation provisions. Use this when the exact citation is unknown or the request is conceptual. Use get_statute_by_citation first for a known INA/USC/CFR citation. Use lookup_criminal_code(jurisdiction='federal'), not this tool, for Title 18 and other federal criminal-code provisions. A separate note_records section lists up to 3 uncodified statutory notes (public-law text printed as notes under U.S. Code sections) whose text matches the query but that are not in records; notes can be spent or superseded, so confirm current effect.

search_statutes

Search unpublished BIA and AAO agency decisions. This lane covers unpublished agency decisions only: unpublished Board removal decisions and AAO administrative appeals. Nonprecedential circuit dispositions are court authority and rank in the ordinary case search -- call search_cases with source='circuit-courts' (or 'circuit-unpublished'), circuit=, and precedential_status='unpublished' for those. Use this for non-binding candidate support after controlling authority is framed, and source-text check candidates with fetch or corpus text tools before relying on them. favorable_only=true restricts results to decisions whose stored direction verifies the immigrant won or DHS/USCIS lost; the tag is derived metadata, so verify direction against source text. For a primary unpublished lane, pass an explicit source; unscoped or inferred-source results are candidates only and must not be treated as completed legal research.

search_unpublished_cases