BrowseAI Dev

Research infrastructure for AI agents — real-time web search, evidence extraction, and structured citations. Every claim is backed by a URL. Every answer has a confidence score.

Agent → BrowseAI Dev → Internet → Verified answers + sources

Website · Playground · API Docs · Discord

---

How It Works

search → fetch pages → extract claims → build evidence graph → cited answer

Every answer goes through a 6-step verification pipeline. No hallucination. Every claim is backed by a real source.

Verification & Confidence Scoring

Confidence scores are evidence-based — not LLM self-assessed. After the LLM extracts claims and sources, a post-extraction verification engine checks every claim against the actual source page text:

1. BM25 sentence matching — Each claim is scored against every sentence in its cited sources using BM25 (the ranking algorithm behind Elasticsearch and Lucene). This catches paraphrased claims that simple keyword overlap would miss.
2. Domain authority scoring — 10,000+ domains across 5 tiers (institutional .gov/.edu → major news → tech journalism → community → low-quality), stored in Supabase with Majestic Million bulk import. Scores self-improve over time using Bayesian cold-start smoothing — every query feeds back verification data to make future scores more accurate.
3. Source quote verification — LLM-extracted quotes are verified against actual page text using hybrid matching (exact substring → BM25 fallback).
4. Cross-source consensus — Each claim is verified against all available page texts (not just cited sources). Claims supported by 3+ independent domains get "strong consensus", boosting confidence. Single-source claims are flagged as "weak".
5. Contradiction detection — Claim pairs are analyzed for semantic conflicts using topic overlap + negation asymmetry. Detected contradictions are surfaced in the response and penalize the confidence score.
6. 7-factor confidence formula — Final score combines: verification rate (25%), domain authority (20%), source count (15%), consensus (15%), domain diversity (10%), claim grounding (10%), and citation depth (5%). Each detected contradiction subtracts 0.05 from the raw score.

Claims include verified, verificationScore, consensusCount, and consensusLevel fields. Sources include verified and authority. Detected contradictions are returned at the top level. Agents can use these fields to make trust decisions programmatically.

Thorough Mode

Pass depth: "thorough" to automatically retry with a rephrased query when first-pass confidence is below 60%. The system searches again with alternative terms, merges sources from both passes, and picks the higher-confidence result.

curl -X POST https://browseai.dev/api/browse/answer \
  -H "Content-Type: application/json" \
  -H "X-Tavily-Key: tvly-xxx" \
  -H "X-OpenRouter-Key: sk-or-xxx" \
  -d '{"query": "What is quantum computing?", "depth": "thorough"}'

Streaming API

Get real-time progress instead of waiting for the full response. The streaming endpoint sends Server-Sent Events (SSE) as each pipeline step completes:

curl -N -X POST https://browseai.dev/api/browse/answer/stream \
  -H "Content-Type: application/json" \
  -H "X-Tavily-Key: tvly-xxx" \
  -H "X-OpenRouter-Key: sk-or-xxx" \
  -d '{"query": "What is quantum computing?"}'

Events: trace (progress), sources (discovered early), result (final answer), done.

Retry with Backoff

All external API calls (Tavily search, OpenRouter LLM, Brave search, page fetching) automatically retry on transient failures (429 rate limits, 5xx server errors) with exponential backoff and jitter. Auth errors (401/403) fail immediately — no wasted retries.

Research Memory (Sessions)

Persistent research sessions that accumulate knowledge across multiple queries. Later queries automatically recall prior verified claims, building deeper understanding over time.

Sessions require a BrowseAI Dev API key (bai_xxx) for identity and ownership. BYOK (Tavily + OpenRouter keys only) works for search/answer but cannot use sessions. Get a free key at browseai.dev/dashboard. For MCP, set BROWSE_API_KEY env var. For Python SDK, pass api_key="bai_xxx". For REST API, use Authorization: Bearer bai_xxx.

# Python SDK
session = client.session("quantum-research")
r1 = session.ask("What is quantum entanglement?")       # 13 claims stored
r2 = session.ask("How is entanglement used in computing?")  # 12 claims recalled!
knowledge = session.knowledge()  # Export all accumulated claims

# Share with other agents or humans
share = session.share()  # Returns shareId + URL
# Another agent forks and continues the research
forked = client.fork_session(share.share_id)

# REST API
curl -X POST https://browseai.dev/api/session \
  -H "Authorization: Bearer bai_xxx" \
  -d '{"name": "my-research"}'
# Returns session ID, then:
curl -X POST https://browseai.dev/api/session/{id}/ask \
  -H "Authorization: Bearer bai_xxx" \
  -d '{"query": "What is quantum entanglement?"}'

# Share a session publicly
curl -X POST https://browseai.dev/api/session/{id}/share \
  -H "Authorization: Bearer bai_xxx"

# Fork a shared session (copies all knowledge)
curl -X POST https://browseai.dev/api/session/share/{shareId}/fork \
  -H "Authorization: Bearer bai_xxx"

Each session response includes recalledClaims and newClaimsStored. Sessions can be shared publicly and forked by other agents — enabling collaborative, multi-agent research workflows.

Query Planning

Complex queries are automatically decomposed into focused sub-queries with intent labels (definition, evidence, comparison, counterargument, technical, historical). Each sub-query targets a different aspect of the question, maximizing source diversity. Simple factual queries skip planning entirely — no added latency.

Self-Improving Accuracy

The entire verification pipeline improves automatically with usage:

• Domain authority — Bayesian cold-start smoothing adjusts domain trust scores as evidence accumulates. Static tier scores dominate initially, then real verification rates take over.
• Adaptive BM25 thresholds — Claim verification thresholds tune per query type based on observed verification rates. Too strict? Loosens up. Too lenient? Tightens.
• Consensus threshold tuning — Cross-source agreement thresholds adapt based on query type performance.
• Confidence weight optimization — The 7-factor confidence formula rebalances weights per query type when user feedback indicates inaccuracy.
• Page count optimization — Source fetch counts adjust based on confidence outcomes per query type.

Feedback Loop

Submit feedback on results to accelerate learning. Agents and users can rate results as good, bad, or wrong — this feeds directly into the adaptive threshold engine.

curl -X POST https://browseai.dev/api/browse/feedback \
  -H "Content-Type: application/json" \
  -d '{"resultId": "abc123", "rating": "good"}'

client.feedback(result_id="abc123", rating="good")
# Or flag a specific wrong claim:
client.feedback(result_id="abc123", rating="wrong", claim_index=2)

Quick Start

Python SDK

pip install browseai

from browseai import BrowseAI

client = BrowseAI(api_key="bai_xxx")

# Research with citations
result = client.ask("What is quantum computing?")
print(result.answer)
print(f"Confidence: {result.confidence:.0%}")
for source in result.sources:
    print(f"  - {source.title}: {source.url}")

# Thorough mode — auto-retries if confidence < 60%
deep = client.ask("What is quantum computing?", depth="thorough")

Framework integrations:

pip install browseai[langchain]   # LangChain tools
pip install browseai[crewai]      # CrewAI integration

# LangChain
from browseai.integrations.langchain import BrowseAIAskTool
tools = [BrowseAIAskTool(api_key="bai_xxx")]

# CrewAI
from browseai.integrations.crewai import BrowseAITool
researcher = Agent(tools=[BrowseAITool(api_key="bai_xxx")])

MCP Server (Claude Desktop, Cursor, Windsurf)

npx browse-ai setup

Or manually add to your MCP config:

{
  "mcpServers": {
    "browse-ai": {
      "command": "npx",
      "args": ["-y", "browse-ai"],
      "env": {
        "SERP_API_KEY": "your-search-key",
        "OPENROUTER_API_KEY": "your-llm-key",
        "BROWSE_API_KEY": "bai_xxx"
      }
    }
  }
}

BROWSE_API_KEY is optional for search/answer but required for Research Memory (sessions).

REST API

# With your own keys (BYOK — free, no limits)
curl -X POST https://browseai.dev/api/browse/answer \
  -H "Content-Type: application/json" \
  -H "X-Tavily-Key: tvly-xxx" \
  -H "X-OpenRouter-Key: sk-or-xxx" \
  -d '{"query": "What is quantum computing?"}'

# Thorough mode (auto-retries if confidence < 60%)
curl -X POST https://browseai.dev/api/browse/answer \
  -H "Content-Type: application/json" \
  -H "X-Tavily-Key: tvly-xxx" \
  -H "X-OpenRouter-Key: sk-or-xxx" \
  -d '{"query": "What is quantum computing?", "depth": "thorough"}'

# With a BrowseAI Dev API key
curl -X POST https://browseai.dev/api/browse/answer \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer bai_xxx" \
  -d '{"query": "What is quantum computing?"}'

Self-Host

git clone https://github.com/BrowseAI-HQ/BrowseAI-Dev.git
cd BrowseAI-Dev
pnpm install
cp .env.example .env
# Fill in: SERP_API_KEY, OPENROUTER_API_KEY
pnpm dev

API Keys

Three ways to authenticate:

Method	How	Limits
BYOK (recommended)	Pass X-Tavily-Key and X-OpenRouter-Key headers	Unlimited, free (search/answer only — no sessions)
BrowseAI Dev API Key	Pass Authorization: Bearer bai_xxx	Unlimited + sessions, sharing, forking
Demo	No auth needed	5 queries/hour per IP

Get a BrowseAI Dev API key from the dashboard — it bundles your Tavily + OpenRouter keys into one key for CLI, MCP, and API use.

Project Structure

/apps/api              Fastify API server (port 3001)
/apps/mcp              MCP server (stdio transport, npm: browse-ai)
/packages/shared       Shared types, Zod schemas, constants
/packages/python-sdk   Python SDK (PyPI: browseai)
/src                   React frontend (Vite, port 8080)
/supabase              Database migrations

API Endpoints

Endpoint	Description
POST /browse/search	Search the web
POST /browse/open	Fetch and parse a page
POST /browse/extract	Extract structured claims from a page
POST /browse/answer	Full pipeline: search + extract + cite. Pass depth: "thorough" for auto-retry
POST /browse/answer/stream	Streaming answer via SSE — real-time progress events
POST /browse/compare	Compare raw LLM vs evidence-backed answer
GET /browse/share/:id	Get a shared result
GET /browse/stats	Total queries answered
GET /browse/sources/top	Top cited source domains
GET /browse/analytics/summary	Usage analytics (authenticated)
POST /session	Create a research session
POST /session/:id/ask	Research with session memory (recalls + stores claims)
POST /session/:id/recall	Query session knowledge without new search
GET /session/:id/knowledge	Export all session claims
POST /session/:id/share	Share a session publicly (returns shareId)
GET /session/share/:shareId	View a shared session (public, no auth)
POST /session/share/:shareId/fork	Fork a shared session into your account
GET /session/:id	Get session details
GET /sessions	List your sessions (authenticated)
DELETE /session/:id	Delete a session (authenticated)
POST /browse/feedback	Submit feedback on a result (good/bad/wrong)
GET /browse/learning/stats	Self-learning engine stats
GET /user/stats	Your query stats (authenticated)
GET /user/history	Your query history (authenticated)

MCP Tools

Tool	Description
browse_search	Search the web for information on any topic
browse_open	Fetch and parse a web page into clean text
browse_extract	Extract structured claims from a page
browse_answer	Full pipeline: search + extract + cite. Supports depth: "thorough"
browse_compare	Compare raw LLM vs evidence-backed answer
browse_session_create	Create a research session (persistent memory)
browse_session_ask	Research within a session (recalls prior knowledge)
browse_session_recall	Query session knowledge without new web search
browse_session_share	Share a session publicly (returns share URL)
browse_session_knowledge	Export all claims from a session
browse_session_fork	Fork a shared session to continue the research
browse_feedback	Submit feedback on a result to improve accuracy

Python SDK

Method	Description
client.search(query)	Search the web
client.open(url)	Fetch and parse a page
client.extract(url, query=)	Extract claims from a page
client.ask(query, depth=)	Full pipeline with citations. depth="thorough" for auto-retry
client.compare(query)	Raw LLM vs evidence-backed
client.session(name)	Create a research session
session.ask(query, depth=)	Research with memory recall
session.recall(query)	Query session knowledge
session.knowledge()	Export all session claims
session.share()	Share session publicly (returns shareId + URL)
client.get_session(id)	Resume an existing session by ID
client.list_sessions()	List all your sessions
client.fork_session(share_id)	Fork a shared session into your account
session.delete()	Delete a session
client.feedback(result_id, rating)	Submit feedback (good/bad/wrong) to improve accuracy

Async support: AsyncBrowseAI with the same API.

Examples

See the examples/ directory for ready-to-run agent recipes:

Example	Description
research-agent.py	Simple research agent with citations
code-research-agent.py	Research libraries/docs before writing code
hallucination-detector.py	Compare raw LLM vs evidence-backed answers
langchain-agent.py	BrowseAI Dev as a LangChain tool
crewai-research-team.py	Multi-agent research team with CrewAI
research-session.py	Research sessions with persistent memory

Environment Variables

Variable	Required	Description
SERP_API_KEY	Yes	Web search API key (Tavily)
OPENROUTER_API_KEY	Yes	LLM API key (OpenRouter)
KV_REST_API_URL	No	Vercel KV / Upstash Redis REST URL (falls back to in-memory cache)
KV_REST_API_TOKEN	No	Vercel KV / Upstash Redis REST token
SUPABASE_URL	No	Supabase project URL
SUPABASE_SERVICE_ROLE_KEY	No	Supabase service role key
PORT	No	API server port (default: 3001)

Tech Stack

• API: Node.js, TypeScript, Fastify, Zod
• Search: Tavily API
• Parsing: @mozilla/readability + linkedom
• AI: Gemini 2.5 Flash via OpenRouter
• Caching: Redis or in-memory with intelligent TTL (time-sensitive queries get shorter TTL)
• Frontend: React, Tailwind CSS, shadcn/ui, Framer Motion
• MCP: @modelcontextprotocol/sdk
• Python SDK: httpx, Pydantic
• Database: Supabase (PostgreSQL)

Community

• Discord — questions, feedback, showcase
• GitHub Issues — bugs, feature requests

Contributing

See CONTRIBUTING.md for setup instructions, coding conventions, and PR process.

License

MIT