Google Surf

google-surf-mcp

English | Korean

Web searches, papers, and GitHub repositories are stored as PKM, ontology, and lineage. The view above is generated with project_memory(action="export", export_format="html", export_view="graph", all_projects=true).

"Turn Google Search, Papers, and Codebases into an Automatic Local Knowledge Graph and lineage for AI Agents with Zero API Key, Zero External Server."

Google Surf stores search and extraction results in a project-scoped local knowledge graph.

As you search, papers, code, web sources, session intent, plans, experiments, and decisions accumulate in a personal PKM. New research searches stored knowledge and fresh web results together, reducing repeated work while continuing to discover new information.

Projects remain isolated by default. Only verifiable links such as matching DOIs, repository URLs, or explicit aliases are added, so knowledge from one project can be reused in another without merging the original records.

Retrieval runs exact search, BM25, vector search, code and graph search, and live web independently, then combines them with RRF and one shared reranker.

Live web + Papers + Codebases + Project memory
                        ↓
Exact + BM25 + Vector + Code graph + Graph PPR
                        ↓
              RRF + Shared reranker
                        ↓
        Results with evidence and provenance

Seven tools are available by default: search / search_parallel / extract / scholar_search / project_memory_search / project_memory / health.

Research mode and automatic capture are enabled by default. Set SURF_RESEARCH=false to use search and extraction without opening the database or graph sidecar; the project memory tools are not registered in that mode.

With research mode disabled, live search and search_parallel results still use a lightweight in-memory reranker. It fuses the provider order with query BM25 rank through RRF without loading the vector model or opening local storage.

Browser search needs no API key. SearchApi can be configured as an optional primary provider or fallback.

Core features

• Web, paper, and codebase search: Google web search is the default, while Scholar is used for paper-specific metadata. SearchApi can act as an optional primary provider or fallback.
• Web and academic document extraction: HTML and PDF extraction returns available titles, authors, DOIs, publication metadata, and body text. search and search_parallel can include abstracts or full bodies.
• Automatic project memory: Search results, extracted bodies, and code repositories are stored in the current project. Unread results remain metadata; extracted content becomes active RAG evidence.
• Structured codebase search: Tree-sitter links files, symbols, imports, and calls from local projects and relevant GitHub repositories. Exact, BM25, vector, and graph search retrieve the code.
• Graph hybrid retrieval: Fresh web results, papers, stored content, codebases, and the project graph are searched independently. Exact, BM25, vector, and PPR candidates are combined through RRF and one shared reranker.
• Ontology and data lineage: Web sources, papers, code, plans, experiments, and decisions become typed entities and relations. Evidence paths run from source through documents, chunks, symbols, evidence, and assertions.
• Cross-project knowledge reuse: Projects remain isolated but selected projects can be searched together. Only verifiable links connect matching entities.
• Durable research history: Session intent, plans, experiments, failures, and decisions supplied by the MCP host are stored as revisions and linked to their supporting evidence.
• Local graph analysis and export: PageRank, PPR, connected components, and Louvain communities run without a separate Neo4j server. Results export to HTML, Graphviz, D3 JSON, and Neo4j import formats.

Search

• API-key-free system Chrome search
• Dedicated logged-out profile that never reads or copies the user's Chrome profile
• Multi-strategy SERP parsing with geometric verification
• Sponsored block and knowledge panel removal
• CAPTCHA detection and environment-specific recovery
• Parser self-healing and context fallback

Numbers

Search speed

Live web search reranker check

• web search responses with BM25+RRF reduce large retrieval payloads by up to 75%. Measured across three uncached queries per provider on a workstation with a 1Gb/s connection. Network and Google response time vary.

Tech Stack

• Runtime: Node.js, TypeScript, Model Context Protocol SDK
• Web search: System Chrome + CDP, Playwright compatibility fallback, SearchApi fallback
• Web extraction: Mozilla Readability, Turndown
• PDF extraction: LiteParse/PDFium, optional OCR, pdf-lib metadata parsing
• Code collection: Local project roots and gated GitHub sparse download
• Code parsing: Tree-sitter for files, symbols, imports, and call relations
• Code search: Exact lookup, BM25, Multilingual E5 vector search, and graph PPR
• Local database: Embedded SurrealDB on RocksDB
• Hybrid retrieval: Live web, papers, project memory, and codebase results combined through RRF
• Ranking: Reciprocal Rank Fusion and shared vector reranking
• Graph analysis: Graphology, PageRank, PPR, connected components, and Louvain communities
• Knowledge model: Versioned ontology, data lineage, cross-project schema and entity linking
• Recovery: CAPTCHA recovery, Playwright pool fallback, and deterministic parser self-healing

Install

Requires Node 20.18.1+. Browser mode also requires Google Chrome or Chromium.

npx google-surf-mcp   # actual MCP - register in client config

First tool call auto-bootstraps the warm profile (you may see Chrome open briefly).

Or local clone:

git clone https://github.com/HarimxChoi/google-surf-mcp
cd google-surf-mcp
npm install

If auto-bootstrap fails (rare), run it manually:

npm run bootstrap

Override paths if needed:

CHROME_PATH=/path/to/chrome SURF_TZ=America/New_York npm run bootstrap

Optional Codex output protection

Google Surf can install an opt-in Codex hook that ranks oversized Bash output before it reaches the model. The host still executes the command; the hook only applies stateless source-order, exact, and BM25 ranking with RRF. It opens neither the Google Surf database nor a browser and does not store command output.

npx -y google-surf-mcp@latest hooks install --host codex

Restart Codex, then open /hooks to review and trust the definitions. Shell output is reranked from 1,500 characters, with a default 1,500-character display budget that expands up to 3,000 for distinct query-matching evidence records or blocks. An explicit override may exceed 3,000; there is no cumulative per-turn output limit. Two identical or near-duplicate searches and explicit foreground polling loops are blocked. Existing write_stdin polls for an already-running unified command remain controlled by the Codex runtime, not this hook.

JSON/JSONL summaries use compact records with input-relative JSON pointers, ancestor identity/condition references and original result values. Markdown table rows retain headers and adjacent context, including source lines. Selected records and their required context must fit together; the renderer reports omissions instead of slicing records. Plain logs retain block reranking. Selection is deterministic and lexical, without a model, database or extra subprocess. Context extraction uses structural/field-name heuristics, not semantic understanding; unselected fields remain in the original input. These excerpts do not replace experiment logs or SSOT history. Input-relative pointers are not persistent retrieval handles. Existing installed hook bundles require an update to use changed code.

npx -y google-surf-mcp@latest hooks status --host codex
npx -y google-surf-mcp@latest hooks update --host codex
npx -y google-surf-mcp@latest hooks uninstall --host codex

Use with Claude Code

Paste this into your ~/.claude.json:

{
  "mcpServers": {
    "google-surf": {
      "command": "npx",
      "args": ["-y", "google-surf-mcp"]
    }
  }
}

Restart Claude Code. All seven tools, including project_memory_search and project_memory, are available by default.

For other MCP clients, use the same JSON shape in their config file.

Search providers

Browser search remains the default. SearchApi can be selected as the primary provider or used only when browser search fails.

SURF_SEARCH_PROVIDER controls search and search_parallel. SURF_SCHOLAR_PROVIDER controls scholar_search. SearchApi modes require your own SearchApi account, key, and available credits.

SURF_BROWSER_ENGINE=auto selects native Chrome on a local desktop and the Playwright compatibility path in cloud or remote-debug mode. Native mode uses a normal hidden Chrome window, not headless Chrome. Set native or playwright to pin the engine.

{
  "mcpServers": {
    "google-surf": {
      "command": "npx",
      "args": ["-y", "google-surf-mcp"],
      "env": {
        "SEARCH_API": "your-searchapi-key",
        "SURF_SEARCH_PROVIDER": "fallback",
        "SURF_SCHOLAR_PROVIDER": "searchapi"
      }
    }
  }
}

Local clone variant:

{
  "mcpServers": {
    "google-surf": {
      "command": "node",
      "args": ["/abs/path/to/google-surf-mcp/build/index.js"]
    }
  }
}

Tools

• search(query, limit?, extract_mode?, extract_limit?, response_content?, max_chars?) - primary single-query tool for live discovery and reading. When new sources must be found and read, set extract_mode in this call instead of downloading PDFs, cloning repositories, or calling extract separately. Use extract only when the exact public URL is already known and no discovery is needed. With project_id, stored project knowledge is fused with live results, but the call never becomes local-only. limit is 1-20. Extraction defaults to none; extract_limit is 1-10 with default 5. response_content defaults to summary to bound one-call output.
• scholar_search(query, limit?) - Google Scholar metadata search, max 10 papers. Supports browser, SearchApi primary, and fallback modes.
• search_parallel(queries[], limit?, extract_mode?, extract_limit?, response_content?, max_chars?) - primary multi-query tool for broad live discovery and reading through a continuous four-tab queue. Set extract_mode in the same call when public web pages, PDFs, papers, or GitHub repositories must be read. Use local PDF tools only for local files or visual layout work, and clone repositories only for editing, building, testing, or full Git history. limit is 1-20 per query. The call-wide extract_limit defaults to 12 and allows up to 20 for abstract; full defaults to and allows 10. response_content defaults to summary to bound one-call output.
• Integrated search extraction reports requested, applied, skipped, truncated, and total_chars. remaining_urls can be passed to extract without repeating the search.
• extract(url, max_chars?, mode?, response_content?) - secondary extraction tool for an exact public URL when no new discovery is required. If sources still need to be found, use search or search_parallel with extract_mode instead.
  • mode="full" (default): reads up to 1000000 characters for research capture. Research mode stores deterministic 4000-character chunks; response_content="full" returns up to 50000 characters and summary returns a 1500-character evidence excerpt.
  • mode="abstract": ~1500-char survey (PDF page 1 or HTML meta description). Document metadata is included and stored with the survey when research mode is enabled.
  • : metadata without body text. Returns available title, authors, publication, dates, DOI, description, keywords, canonical URL, and PDF properties including page count.

Replayable research collection

google-surf-collect runs a versioned JSON specification through one persistent MCP session. A specification can mix live search jobs with local-only project_memory_search jobs. Live jobs can extract bodies in the same call, while local jobs reuse indexed project knowledge without opening Google.

npx google-surf-collect examples/research-collection.example.json

From a source checkout:

npm run build
npm run research:collect -- examples/research-collection.example.json

A project workflow can also record durable sessions and plans, rebuild approved code roots, search the resulting local knowledge, and export its graph:

npm run research:collect -- examples/project-memory-workflow.example.json

project_memory collection jobs allow record, rebuild, and export. Destructive forget operations are not accepted by the collection schema. Project-level project_id is inherited by every job unless an all-project export is requested.

The output is append-only JSONL. Its manifest records the normalized specification hash, package version, Git commit, Node runtime, platform, project setup, and server health. Every search, record, rebuild, and export result records the stable job id, exact tool arguments, attempt, timestamps, elapsed time, response, and error state. Successful jobs are skipped on resume; failed jobs are retried. A changed specification requires a new output file. Set project_name with project_id when the runner should create a missing project; existing projects are reused.

Set retrieval_mode to live when prior project RAG state must not affect live result ranking. Results are still captured under project_id. Use hybrid when the collection intentionally ranks new web evidence together with stored project knowledge. API keys and environment variable values are never written to the collection log.

This makes the collection procedure and returned snapshot replayable and auditable. Live web results can still change with time, locale, network route, and upstream ranking.

Graph hybrid RAG with ontology and lineage

flowchart TB
    subgraph SOURCES["1. Search and research"]
        direction LR
        LIVE["Live web<br/>Google browser • SearchApi fallback"]
        PAPER["Pages and papers<br/>extract • Scholar metadata"]
        PROJECT_INPUT["Code and project records<br/>local roots • GitHub • host-provided session/plan"]
    end

    INGEST["2. Deterministic capture<br/>normalize • deduplicate • content hash<br/>repository source gate • Tree-sitter"]

    subgraph KNOWLEDGE_BASE["3. SurrealDB knowledge base"]
        direction LR
        CONTENT["Content and code index<br/>exact • BM25 • HNSW vector<br/>documents • chunks • symbols"]
        PROV["Data lineage and provenance<br/>source → evidence → assertion<br/>valid time • recorded time • correction"]
        ONTOLOGY["Versioned ontology<br/>core/project term revisions<br/>entity type • relation • alias • merge/split"]
        MEMORY["Project memory<br/>session intent • plan revisions<br/>experiments • decisions"]
    end

    subgraph INTELLIGENCE["4. Graph intelligence"]
        direction LR
        SCHEMA["Cross-project schema linking<br/>type and relation alignment<br/>stable identifier → identity bridge"]
        SIDECAR["Typed graph sidecar<br/>PageRank • Louvain • query-time PPR"]
    end

    FUSION["5. Hybrid retrieval<br/>live • exact • BM25 • vector • graph<br/>deterministic RRF • shared reranker • fresh-web floor"]
    RESULTS["Results + provenance<br/>short storage receipt"]

    LIVE --> INGEST
    PAPER --> INGEST
    PROJECT_INPUT --> INGEST
    INGEST --> CONTENT
    INGEST --> PROV
    INGEST --> ONTOLOGY
    INGEST --> MEMORY
    ONTOLOGY --> SCHEMA
    CONTENT --> SIDECAR
    PROV --> SIDECAR
    MEMORY --> SIDECAR
    SCHEMA --> SIDECAR
    LIVE --> FUSION
    CONTENT --> FUSION
    SIDECAR --> FUSION
    FUSION --> RESULTS
    RESULTS -. "automatic search/extract capture" .-> INGEST

    classDef inputStyle fill:#eef6ff,stroke:#2563eb,color:#172554
    classDef processStyle fill:#fff7ed,stroke:#ea580c,color:#431407
    classDef storageStyle fill:#ecfdf5,stroke:#059669,color:#052e16
    classDef intelligenceStyle fill:#f5f3ff,stroke:#7c3aed,color:#2e1065
    classDef outputStyle fill:#f8fafc,stroke:#475569,color:#0f172a
    class LIVE,PAPER,PROJECT_INPUT inputStyle
    class INGEST processStyle
    class CONTENT,PROV,ONTOLOGY,MEMORY storageStyle
    class SCHEMA,SIDECAR intelligenceStyle
    class FUSION,RESULTS outputStyle

One local knowledge base

One SurrealDB instance is the authority for three distinct layers:

• Catalog: stable identities, revisions, project membership, ontology, provenance, and temporal links.
• Payloads: exact bodies, source snapshots, manifests, and artifact references. Identical content is stored once and can be fetched by ID and byte span.
• Derived views: exact, BM25, vector, code, and graph indexes plus the compact current-project view. These are versioned by source hash and can be rebuilt.

A single local research broker owns the embedded RocksDB connection. Multiple MCP sessions connect through authenticated local IPC, run bounded reads concurrently, and serialize writes without opening another database process. Original evidence and historical occurrences remain addressable even when bodies, embeddings, or graph structures are deduplicated.

Durable records and incremental indexing

project_memory writes project profiles, plans, experiments, attempts, measurements, artifacts, documents, decisions, and sessions as typed records. A stable asset ID receives revisions; an uncertain retry with the same idempotency key returns the committed revision with idempotent_replay=true and applied_this_request=false. body_path streams bodies up to 256 MiB, artifact manifests are paginated instead of sliced, and get/get_batch return exact typed or legacy records through bounded body spans. Compact write receipts report accepted bytes, references, revision, and the exact readback handle without echoing the submitted body.

sync accepts registered roots or explicit changed and removed paths. Unchanged roots and parent identities remain intact, a no-op does not rewrite derived indexes, and a one-file change publishes only the affected generation. Code, retrieval, and graph publication run under one durable job ID. job_wait uses a revision cursor, job_cancel stops at a safe stage boundary, and unfinished jobs resume after process restart. Under the worker soft-memory limit, foreground requests finish first; rebuildable index jobs checkpoint at a durable stage boundary and resume in the replacement worker without cancelling the committed record.

Ontology and cross-project links

The versioned ontology preserves entity type and relation changes as revisions. Schema linking aligns project-specific types and relations with the shared schema. Entity linking connects the same paper, repository, or entity only when verifiable identifiers such as a DOI, repository URL, or explicit alias match. Ambiguous candidates are not linked automatically.

Data and research lineage

• Source lineage: source → document → chunk → evidence → assertion
• Code lineage: repository → directory → file → symbol → import/call
• Research lineage: session → intent → plan revision → experiment → decision

This preserves the evidence behind claims and decisions while retaining corrected or superseded history.

Graph retrieval

Graphology builds a typed graph projection from SurrealDB and computes PageRank, connected components, and Louvain communities. At retrieval time, related nodes seed PPR-based multi-hop search. Live web, exact, BM25, vector, and graph candidates are combined through deterministic RRF and one shared reranker.

Local multi-query retrieval batches query embeddings, keeps exact, BM25, and vector lanes independent through RRF, expands the graph once, hydrates selected chunks once, and reranks once. Graph-only all-project searches use a lightweight memory-node index and verified identity aliases to select at most four graph scopes instead of constructing every project graph at query time.

If query embedding fails or returns no vector, local search returns its lexical/graph ranking without retrying the model in reranking. Timing fields are non-overlapping wall-clock partitions: embedding_ms plus retrieval_ms reaches the end of retrieval, and rerank_ms covers the remaining response work. Retrieval may overlap embedding, so retrieval_ms is not standalone database execution time. A supplied response_deadline_ms starts after broker admission; the transport watchdog retains its five-minute queue/IPC allowance in addition to that deadline. Native queries that outlive a partial response retain their broker read slot until they settle.

Research retrieval check

The 60-query regression fixture compares the lexical baseline with the complete local RAG path. To separate retrieval changes from HNSW approximation, the check pins one embedding thread and uses exhaustive vector scoring. The table reports the median of three fresh-process runs:

All retrieved targets passed exact provenance readback. The conservative reranker preserved the fused order on this fixture; this check does not claim production search quality or a reranker gain.

npm run research:vector-backends -- --rows 5000 compares physical representations with deterministic test vectors. The sample below excludes E5 model loading and should be rerun on the target machine:

The default remains HNSW for full-chunk semantic coverage. Compact indexes one representative chunk per source; it used less memory and matched aggregate recall in this short-record fixture, but can miss rare details near the tail of long sources.

Broker and durability check

npm run research:reopen-probe also verifies byte-exact body paging, a paginated 225-reference artifact manifest, and a pending rebuild that resumes after process restart and becomes searchable. The soak runs with the vector model disabled to isolate broker, database, and query lifecycle memory; vector backends are measured separately above.

Project isolation and knowledge reuse

project_id selects where new results are stored. include_project_ids expands the read scope without changing the write target. Original records remain isolated by project, while verified schema and entity links allow papers, code, and experiment results to be reused across selected projects.

Interactive graph and export

Use project_memory(action="export", export_format="html", export_view="graph"). The returned standalone HTML opens locally without a server and contains three coordinated views. Use project_id for one project, include_project_ids for a selected combined graph, or all_projects=true for every project.

• PKM groups the integrated project graph by community and sizes nodes by PageRank.
• Lineage separates source and code lineage from session, intent, plan, experiment, and decision lineage while keeping both flows aligned by stage.
• Ontology shows core types and relations, aligned shared schema, and typed instances. A verified identity layer appears only when stable identifiers or explicit aliases prove a cross-project match.

Search, type filters, one to three hop local focus, pan, zoom, and the provenance inspector work inside the file. Large graphs use a deterministic semantic projection that balances node type, PageRank, degree, and community coverage. The viewer reports source and displayed counts, replaces internal IDs with local aliases, disambiguates repeated labels, and does not embed source IDs, local paths, node bodies, plan text, or evidence quotes.

The project menu switches between every project embedded in the export and an integrated All projects view. Use all_projects=true when exporting to include every named project in the local database, or use project_id and include_project_ids for a bounded set. Clicking empty canvas space clears local node focus. PNG exports the current canvas, while JSON exports the current tab, project, type filters, and local focus using only the anonymized viewer payload. The standalone file has a nonce-bound script CSP, makes no network connections, and permits source links only for stripped public HTTP or HTTPS URLs.

Neo4j export

Use project_memory(action="export", export_format="neo4j", export_view="graph"). The returned directory contains nodes.csv, relationships.csv, constraints.cypher, load.cypher, manifest.json, and README.txt. PageRank, community, ontology, lineage, project IDs, source IDs, and evidence IDs are preserved. Node bodies, plan text, and evidence quotes are not exported.

For a new or empty local database, run the Neo4j offline importer from the export directory:

neo4j-admin database import full --nodes=nodes.csv --relationships=relationships.csv neo4j

For an existing local database, copy both CSV files to the Neo4j import directory, then run:

cypher-shell -f constraints.cypher
cypher-shell -f load.cypher

The offline importer creates typed node labels and relationship types. The online loader uses SurfNode and SURF_RELATION, retaining the original kinds and relationship types as properties. neo4j-admin database import full is intended for a new or empty database; use LOAD CSV for an existing database. See the official Neo4j import and LOAD CSV documentation. Bolt is a connection protocol, not an export file format, so this command does not connect to or modify a Neo4j server.

Storage scope and security

Research mode is enabled by default. search, search_parallel, scholar_search, and extract results are captured automatically. Session intent, plans, experiments, and decisions are stored only when the MCP host sends them through project_memory; versioning, ontology mapping, and lineage linking then run automatically. Retrieval mode is server configuration, not a per-call argument.

Image retrieval, image embeddings, and visual reranking are not part of research memory. OCR is used only to recover searchable text from scanned PDF pages.

Credential and private-key files are excluded from body indexing. HTML exports omit source IDs, local paths, node bodies, plan text, and evidence quotes. The viewer initiates no network requests and opens stripped public HTTP or HTTPS source links only after user action.

Project and assertion deletion require a count preview and confirmation token. They create reversible tombstones and preserve evidence and correction history. Fact correction takes only target_id, replacement, and reason; the prior assertion remains as bitemporal history. Plan revisions are append-only. Experiments are bound to the active revision and must be finished explicitly as success, failed, or inconclusive. Logs are not used to infer an outcome. Receipts list stored categories only:

Project: Graph memory | Session: temporal graph research | Stored: paper 1 (Graphiti), repo 1 (getzep), search summaries 3 | Status: ready

Set SURF_RESEARCH=false to keep the database and sidecar closed and omit project_memory_search and project_memory. Obsidian and Notion sync are not included and will remain project-level opt-in when added.

Local operations CLI

These commands use the same authenticated research broker as MCP sessions. They do not open another embedded database owner or browser:

npx google-surf-mcp doctor --json
npx google-surf-mcp daemon start --background --no-window
npx google-surf-mcp memory get --project PROJECT --record RECORD --json
npx google-surf-mcp memory sync --project PROJECT --manifest changes.json --json
npx google-surf-mcp memory current --project PROJECT --json
npx google-surf-mcp job wait --project PROJECT --id JOB --after-revision 0 --json
npx google-surf-mcp job cancel --project PROJECT --id JOB --reason "operator request" --json
npx google-surf-mcp daemon drain --json

doctor is passive: it reports process identity, versions, capabilities, queue counters, memory counters, cache bytes, current lifecycle state, foreground/background ownership, checkpoint-drain state, and the exact blockers preventing worker recycling without opening the database, vector model, or graph. The broker drains after meaningful work is idle even if an MCP client remains connected; active requests and commits postpone the drain, while rebuildable background jobs checkpoint at safe stage boundaries under memory pressure. The next request reconnects transparently.

Env vars

Embedded resource budget

Embedded storage defaults to a 512 MiB RocksDB block cache, a 128 MiB HNSW vector cache, two 32 MiB write buffers, and two background jobs instead of machine-scaled server defaults. Explicit SURREAL_HNSW_CACHE_SIZE, SURREAL_ROCKSDB_BLOCK_CACHE_SIZE, SURREAL_ROCKSDB_WRITE_BUFFER_SIZE, SURREAL_ROCKSDB_MAX_WRITE_BUFFER_NUMBER, SURREAL_ROCKSDB_THREAD_COUNT, and SURREAL_ROCKSDB_JOBS_COUNT settings take precedence.

Background indexing runs one job at a time. Queued captures in the same project share a current indexing pass, and pending capture requests survive broker restarts. Inference calls share one queue, and each small embedding batch is saved before checking for drain. Document indexing does not load unrelated code bodies. Existing cached embeddings remain reusable; the smaller batch setting does not force a full reindex. These are component budgets, not a hard whole-process RSS cap. Native storage defaults apply when a broker opens its database, not to an already running broker.

OrcaRouter

SURF_LLM_HEAL=true
SURF_LLM_PROVIDER=orcarouter
ORCAROUTER_API_KEY=...
SURF_LLM_MODEL=orcarouter/auto

Troubleshooting

• Native search keeps the current session open and shows Chrome when a CAPTCHA appears. Solve it in that window and retry; the next call verifies the page, restores the hidden window guard, and continues with the same session. SearchApi fallback remains available through SURF_SEARCH_PROVIDER=fallback and SURF_SCHOLAR_PROVIDER=fallback.
• Playwright CAPTCHA recovery has 4 modes (picked automatically from env):
  • default (local desktop): OS notification fires, headed Chrome opens, and the call returns; solve it and retry
  • SURF_HEADLESS=false: headed Chrome opens without a notification; solve it and retry
  • SURF_REMOTE_DEBUG=true: DevTools port + instructions printed, attach chrome://inspect locally to solve
  • SURF_CLOUD_MODE=true: fail-fast with CAPTCHA_REQUIRED error
• Headed Chrome opens to a plain search box instead of CAPTCHA: just type any query in the box and press Enter. Subsequent calls work.
• "Chrome not found": install Chrome or set CHROME_PATH.
• Stale selectors: runtime per-strategy reorder (SURF_SELF_HEALING, deterministic) plus a manually dispatched repair workflow (SURF_LLM_HEAL optional, human review required, never auto-merged).
• Playwright searches feel slower than expected: check health().pool.fallback. true means the worker pool is using a single context. Native search uses one authenticated local browser broker across MCP sessions. The broker keeps one hidden Chrome process with up to four reusable tabs for search, search_parallel, and scholar_search. Query starts are staggered. A CAPTCHA shows and preserves that session for user recovery, then the window guard is restored on the next call. A browser crash starts a new session on the next call.
• SSRF: extract blocks localhost, private IPs, AWS metadata by default. Set SURF_ALLOW_PRIVATE=true to allow them.
• Cache cleanup: npm run cache:clear removes search/extract and downloaded vector-model caches. It does not remove the research DB.
• The local research DB is not application-encrypted. Use OS account permissions and disk encryption such as BitLocker or FileVault when the machine or backups need at-rest protection.

Changelog

See CHANGELOG.md.

License

MIT