The risk layer agents can't skip.
Pre-transaction risk infrastructure for AI agents operating on-chain. bitfence intercepts transactions before execution, scores them across 30+ signals in under 50ms, and returns an actionable recommendation.
When an AI agent is about to transact on-chain, bitfence intercepts the intent and evaluates it — before anything is signed. The agent pays $0.003 USDC via x402 and gets back a score, a recommendation, and a reason.
Behind that score: a weighted composite across 6 risk categories — contract authority, liquidity depth, holder distribution, trading patterns, cross-chain deployer reputation, and ML behavioral analysis. Each category evaluates multiple on-chain signals.
Circuit breakers override the composite when critical signals fire — certain flags force a minimum score regardless of what other categories say. Established tokens like USDC are scored differently from new launches with the same flags. Deployer reputation is tracked across chains — a deployer who rugged previous tokens on Solana or across EVM chains gets flagged on their next launch.
Three modes: standard risk scores any token. Contextual risk factors in the agent's position size and portfolio — slippage, MEV exposure, max safe position. Quick check returns a fast cached score for known tokens.
Agent "Check if wrapped SOL is safe before swapping." → GET api.bitfence.ai/v1/risk/solana/So111...1112 x402 payment: $0.003 USDC on Base ✓ settled · tx 0x6e7c...881e · 4012ms ╭──────────────────────────────────────╮ │ Risk Score 12/100 █░░░░░░░░░ │ │ Risk Level LOW │ │ Recommendation PROCEED │ │ Confidence 92% │ ╰──────────────────────────────────────╯ 6 risk categories · 11 signals evaluated Token: SOL · Solana · established → Agent proceeds with the swap.
Agent "Swap 500 USDC for BRETT on Base." → GET api.bitfence.ai/v1/risk/base/0x532f...42E4 x402 payment: $0.003 USDC on Base ✓ settled · tx 0xa41f...917 · 7934ms ╭──────────────────────────────────────╮ │ Risk Score 80/100 ████████░░ │ │ Risk Level HIGH │ │ Recommendation AVOID │ │ Confidence 85% │ ╰──────────────────────────────────────╯ ⚠ Flags triggered: · Mint authority active · Freeze authority active · LP unlocked 6 risk categories · 12 signals evaluated Token: BRETT · Base · not established → Trade blocked. Agent will not execute.
Agent "Buy $5,000 worth of token 7GCi...W2hr on Solana. My portfolio is $25,000." → POST api.bitfence.ai/v1/risk/contextual x402 payment: $0.005 USDC on Base ✓ settled · tx 0xb88c...2f1a · 6211ms ╭──────────────────────────────────────╮ │ Risk Score 52/100 █████░░░░░ │ │ Risk Level MEDIUM │ │ Recommendation REDUCE_SIZE │ │ Confidence 72% │ ╰──────────────────────────────────────╯ Position context: Slippage 1.33% Effective cost $5,067 Max safe position $7,500 MEV exposure medium Portfolio conc. 20% Pool liquidity $4.56M → Agent reduced position to $7,500.
Catch the intent
The agent framework sends the transaction intent to bitfence before anything is signed. Works as middleware — agents don't need to change their logic.
→Evaluate across 30+ signals
30+ on-chain signals across 6 risk categories — evaluated, weighted, and scored into a composite 0–100 in under 50ms.
→Act on the recommendation
The agent receives a score, a risk level, a suggested max position size, and a plain-language reason. Proceed, reduce, or block — the agent decides.
Add risk checks to your agent framework, or connect via MCP. Both paths return the same risk score — choose what fits your stack.
Framework integration
Add bitfence as an ActionProvider. The agent automatically checks risk before any on-chain action. Listed on Coinbase Bazaar for auto-discovery by x402 agents.
import { AgentKit } from "@coinbase/agentkit"; import { bitfenceActionProvider } from "@bitfence/agentkit"; const agent = await AgentKit.from({ walletProvider, actionProviders: [ bitfenceActionProvider(), // your other providers... ], }); // Agent now checks bitfence before // every swap, transfer, or DeFi action. // Pays $0.003 USDC via x402 per query.
Direct access via MCP
Connect any MCP client — Claude Desktop, Cursor, or your own agent — to the bitfence MCP server. Three tools: analyze_token, batch_analyze, and quick_check.
// claude_desktop_config.json { "mcpServers": { "bitfence": { "command": "npx", "args": ["@bitfence/mcp-server"], "env": { "BITFENCE_API_URL": "https://api.bitfence.ai" } } } } // → tools: analyze_token, batch_analyze, // quick_check
Shipping and expanding. Here's what's live today.
bitfence isn't a wallet pop-up or a browser extension. It's infrastructure designed for AI agents that operate without human oversight — where every millisecond and every fraction of a cent matters.