
PumpFun Claims Bot
io.github.nirholas由 GitHub 帳號 nirholas 發布v1.0.1更新於 Oct 25, 2026
MCP server for PumpFun on-chain intelligence — tokens, claims, holders, and bundles.
概覽
本機 MCP 伺服器,為助理提供唯讀的 PumpFun 鏈上情報:代幣資訊、持有人、交易、流動性、捆綁偵測、建立者資料、GitHub 使用者與領取紀錄。
- 可以做什麼
- 查詢代幣中繼資料、市值、聯合曲線進度與標記,以及已畢業代幣的 PumpSwap 池流動性和捆綁偵測(詐騙指標)。檢視前幾大持有人與集中度指標,以及近期交易活動、成交量和買賣筆數。檢視建立者的發幣歷史、詐騙估計和近期代幣,並依使用者名稱或數字 ID 取得 GitHub 資料。查詢某個 GitHub 使用者在 PumpFun 的領取歷史(次數與已領取的代幣)並取得目前 SOL/USD 價格。
- 適用情境
- 適合讓助理在對話中研究 Solana 上的 PumpFun 代幣及其建立者,例如查看持有人、流動性、捆綁情況、建立者歷史,或某個 GitHub 使用者是否領取過社交費用。
- 執行需求
- 透過 npx 或全域 npm 安裝後在本機執行,需要 Node.js 20 或更新版本。鏈上查詢使用 Solana RPC HTTP 位址,預設為公開主網,可能被限流,建議使用專用 RPC。GITHUB_TOKEN 為選用,可將 GitHub API 限額從每小時 60 次提高到 5000 次。需要存取 Solana RPC 與 GitHub 的網路。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 PumpFun Claims Bot,將其新增到工作區。
- 為需要使用其工具的對話啟用該服務。
Desktop only,透過 STDIO。 STDIO 服務會啟動本機處理程序,因此需要 SourceWeft 桌面主機。
其他 MCP 客戶端
參照 儲存庫 中的啟動說明。
README
PumpFun Claims Bot
[PumpFun Claims Bot — Solana on-chain events → enrichment pipeline → Telegram channel cards]
[CI] [License] [Stars] [Telegram] [npm] [MCP] [Node.js] [Solana] [Deploy on Railway]
Read-only Telegram channel feed that broadcasts PumpFun on-chain activity — GitHub social fee claims, token graduations, and more. Posts rich, intelligence-enriched cards to a Telegram channel in real time.
Website · Live Channel · MCP Server · Deploy Guide · Contributing
Looking for interactive monitoring? The telegram-bot supports watch management, group chats, REST API, SSE streaming, and webhooks. Use this channel-bot for simple broadcast-only channels.
Related Source
Pump.fun GitHub Claims — nirholas/pumpfun-github-claims
is a focused companion repository for the @pumpfunclaims
feed. It provides standalone bot source and documentation for first reward
claims per GitHub developer and coin, with previous claimed coins as context.
Its implementation status
tracks the remaining coin-attribution work. This repository offers the broader
toolkit, including web dashboards, an MCP server, credibility scoring and
developer reputation tracking. Both repositories remain available as sources.
Features
What a GitHub claim actually proves
A claim card starts with a real on-chain withdrawal from a Pump.fun GitHub social-fee account. That is useful evidence, but it is not automatically proof that the GitHub user created, owns, or endorses a coin. Token metadata is user-supplied, lookalike accounts exist, and one social-fee account can be shared by several coins.
“First-ever” refers to the GitHub fee account, not necessarily the displayed coin. Trade buttons are shown only for verified repository or creator-wallet relationships. The feed is discovery evidence, never an endorsement.
Feed Types
Claim Intelligence
Every GitHub social fee claim card includes:
Graduation Cards
Rich graduation cards include creator profile, top holders analysis, 24h trading volume, dev wallet activity, pool liquidity, and bundle detection.
Influencer Tier System
Combines GitHub and X/Twitter follower data to classify claimers:
Either threshold triggers the tier — a user with 50K X followers and 50 GitHub followers still qualifies as Influencer.
Credibility Score & Dev Track Record
Every claim card leads with a deterministic 0-100 credibility verdict synthesised from the trust signals the bot already gathers, plus the claiming dev's persistent track record across prior tokens.
The score is deterministic and transparent — same inputs always produce the same number, no model call, and every point is attributed to a named factor shown under the headline:
Scoring starts at a neutral 50 and adjusts by weighted factors: claim verification (+22, or -20 on a GitHub-owner mismatch), account age (+14 for 5y+ down to -18 for < 30 days), public repos, followers, starred non-fork repo, copycats (-10/-16), bundling, holder concentration, prior rugs, and bans.
Dev track record turns that one-shot score into memory. Every score is recorded against the claiming GitHub user id and persisted, so the next time that dev launches a token the card shows their history — exposing a serial fee-farmer whose newest coin looks clean, and crediting a builder with a real record:
Logic lives in src/credibility.ts (pure scoreCredibility) and src/dev-reputation.ts (persistent store), both fully covered by tests.
AI-Powered Claim Summaries
First-time claims get a one-line AI take generated by Groq (llama-3.3-70b-versatile):
- Analyzes 30+ signals: token metadata, GitHub profile, creator history, trading activity
- Max 120 characters — direct, opinionated, trader-focused
- 5-second timeout with graceful fallback (empty string on failure)
- HTML-sanitized output safe for Telegram
Examples:
"Real GitHub project, dev claimed fast" · "Fork of popular repo, proceed with caution"
Architecture
Multi-RPC Failover
The RpcFallback class manages multiple Solana RPC endpoints with automatic rotation:
- Round-robin rotation after 3 consecutive failures on any endpoint
- 60-second cooldown per failed endpoint before retrying
- Earliest-expiry fallback — if all endpoints are in cooldown, picks the one expiring soonest
- Success resets — a single successful call resets the failure counter
- Configure via
SOLANA_RPC_URLS(comma-separated)
Programs Monitored
Claim instructions decoded (v1 and the V2 variants from the October 2026 upgrade):
sweep_creator_fee and sweep_protocol_fee (Pump and PumpAMM) are recognised but are never counted as a claim: a sweep only moves fees the curve or pool still holds into the creator vault, and pays nobody. A claim transaction usually carries a sweep first, then the claim; only the claim (vault to creator) is reported, and each claim is paid by its own event, so nothing is counted twice. A sweep-only transaction posts nothing.
Event decoders read the fields they need from the known prefix and ignore trailing bytes, so both pre-upgrade events and the longer October 2026 layouts decode. Trades recognise the buy_v3, sell_v3, buy_exact_quote_in_v3 and multi_hop_swap instruction names. Since the upgrade a buy can complete the bonding curve and keep buying on the new PumpSwap pool in the same instruction (synthetic migration, before migrate runs); the whale feed adds that PostCompleteBuyEvent amount to the buyer's total and marks the trade as having completed the curve.
Install
npm (MCP server — recommended for AI assistants)
Run the MCP server instantly with npx — no clone needed:
Or install globally:
Add to your MCP client config (Claude Desktop, Cursor, VS Code Copilot):
npm package: pumpfun-claims-bot on npm
From source (full bot + Telegram feed)
Quick Start
1. Create a Telegram Bot
- Message @BotFather on Telegram
/newbot→ follow prompts → copy the bot token- Create a public channel (e.g.,
@pumpfunclaims) - Add the bot as an admin to the channel (must have "Post Messages" permission)
2. Configure Environment
Environment Variables Reference
3. Run
4. Deploy with Docker
5. Deploy to Railway
Railway auto-deploys from GitHub and provides persistent volumes for claim tracking data.
See railway.json for the deployment config:
Project Structure
How It Works
Claim Detection Pipeline
SocialFeeIndex Bootstrap
On startup, the bot fetches all SharingConfig accounts from the PumpFees program to build a reverse mapping from social fee PDA addresses to token mints. This enables resolving which token a social fee claim belongs to without additional RPC calls.
- ~148K mappings loaded at startup
- Incremental updates via WebSocket subscription on
CreateFeeSharingConfigandUpdateFeeSharesevents - Lookup:
socialFeeIndex.getMintForPda(pdaAddress)→ token mint - One PDA → many mints — handles scammers who reuse PDAs across multiple tokens
Fake Claim Detection
Some users call the claim_social_fee_pda instruction targeting random token PDAs where they have no fees to collect. The bot detects these by checking:
- The instruction discriminator matches
claim_social_fee_pdaorclaim_social_fee_pda_v2 - The transaction logs contain no
SocialFeePdaClaimedevent — OR the event showsamountLamports = 0 - The GitHub user ID and platform are still parsed from the instruction args (Anchor Borsh format)
Fake claims are posted with a ⚠️ FAKE CLAIM warning and a 🚩 Fake claim — no fees paid out trust signal.
First-Claim Tracking
The ClaimTracker maintains a persistent set of GitHub user IDs that have successfully claimed:
- In-memory set for fast lookup during processing
- Debounced disk persistence (5-second delay) to
data/github-first-claims.json - Split check/mark pattern:
hasGithubUserClaimed()checks without side effects,markGithubUserClaimed()only called after successful Telegram post - Claim counter:
incrementGithubClaimCount()returns sequential claim number per user - First-claim status is NOT set for fake claims
Three-Layer First-Claim Verification
- Local dedup — Skip if already posted (survives restarts via persisted JSON)
- On-chain verification —
lifetimeClaimedLamports == amountLamportsconfirms it's truly the first claim on-chain - Graceful fallback — Skips first-claim banner if verification fails (prevents false positives after redeployment)
Example Claim Card
Requirements
- Node.js >= 20.0.0
- Telegram bot token (via @BotFather)
- Telegram channel with the bot added as admin
- Solana RPC endpoint — dedicated RPC recommended (Helius, QuickNode, Triton). Public mainnet works but may rate-limit.
- GitHub token (optional) — raises API rate limit from 60 to 5,000 req/hr
Troubleshooting
Bot Not Posting Messages
- Check bot permissions — The bot must be an admin in the channel with "Post Messages" permission
- Verify CHANNEL_ID — Use
@channel_namefor public channels or the numeric ID (e.g.,-100xxx) for private channels. To find the numeric ID, forward a channel message to @userinfobot - Telegram 403 error — Means the bot is NOT a member/admin of the channel. Add it via channel settings → Administrators → Add Administrator
- Check logs — Set
LOG_LEVEL=debugto see all events the bot processes
Rate Limiting
Telegram limits bots to ~30 messages per second to a channel. The grammY framework handles rate limiting automatically:
- Messages may be delayed but won't be dropped
- The bot includes a retry helper that respects
retry_afterheaders - For very high activity, increase
POLL_INTERVAL_SECONDSto reduce event volume
RPC Connection Issues
- Public RPC endpoints have rate limits — for production use a dedicated RPC
- Set
SOLANA_RPC_URLSwith multiple endpoints for automatic failover - If WebSocket disconnects, the bot falls back to HTTP polling at
POLL_INTERVAL_SECONDS - The
RpcFallbackclass provides round-robin across configured endpoints - Set
LOG_LEVEL=debugto see connection status
Missing Claims
- GitHub claims only? — Set
REQUIRE_GITHUB=trueto only post GitHub social fee claims - Feed disabled? — Verify
FEED_CLAIMS=trueis set - SocialFeeIndex slow? — Initial bootstrap fetches ~148K accounts. This takes 30-60 seconds on startup. Check logs for
SocialFeeIndex: loaded N mappings - Rate limited? — GitHub API allows 60 req/hr unauthenticated. Set
GITHUB_TOKENfor 5,000 req/hr
Pipeline Stats
The bot logs pipeline counters every 60 seconds:
- total: All claim events received
- social: GitHub social fee PDA claims
- first/repeat: First-time vs. returning claimers
- posted: Successfully posted to Telegram
- skip cashback: Cashback claims (user refunds, not creator activity)
Web Dashboard
Two React frontends are checked in, and they are not the same app:
The features listed below describe web/web/. Both directories build clean today;
consolidating them onto one is still open work.
Deep links (/dashboard, /docs, ...) are client-side routes, so any static host
must rewrite unknown paths to index.html. Each app now ships a vercel.json with
that rewrite; on a non-Vercel host, configure the equivalent SPA fallback.
Pages
Dashboard Features
- Server-Sent Events (SSE) — Real-time streaming via
/api/v1/claims/streamwith auto-reconnect (3s delay) - Event Filters — All, Launches, Whales, Graduations, Claims, Distributions
- Live Stats Bar — Counters for each event type
- Watch Lists — Add/remove wallet addresses to monitor (
GET/POST/DELETE /api/v1/watches) - Rich Event Cards — Token launches, whale trades, graduations, fee claims with full context
- Connection Status — Visual indicator with error messages
- Mock Data Fallback — Simulated feed when SSE is unavailable
Tech Stack
MCP Server
The bot includes a built-in Model Context Protocol (MCP) server that lets AI assistants (Claude, Copilot, Cursor, etc.) query PumpFun on-chain data conversationally.
MCP Tools
Usage: Stdio Transport (Claude Desktop / Cursor / VS Code)
Add to your MCP client config (e.g. claude_desktop_config.json):
Or run from source:
Usage: Streamable HTTP Transport (Embedded)
Run alongside the main bot by setting MCP_ENABLED=true:
The MCP endpoint is available at POST /mcp on the configured port. Clients connect using the Streamable HTTP transport.
Example Queries
Once connected, ask your AI assistant:
- "Look up token info for mint address 7xKXt...p3Bz"
- "Has GitHub user 12345 ever claimed PumpFun fees?"
- "Who are the top holders of this token?"
- "Check if this token launch was bundled"
- "What's the current SOL price?"
Health Check API
The bot exposes an HTTP health check server for Railway / Docker probes.
Response:
- Returns 200 for
ok, 503 fordegraded - Dynamic stats injected via callback (pipeline counters, connection status)
- Port configurable via
PORTorHEALTH_PORTenv vars (default: 3000)
Testing
The project uses Vitest: 204 tests across 12 suites, all green on npm test.
Test Suites
Every suite lives in src/__tests__/ (plus fixtures.ts, shared sample data).
Local Development
Set LOG_LEVEL=debug — all events are logged to stdout regardless of whether they're posted to Telegram.
Tech Stack
Contributing
Contributions are welcome! See CONTRIBUTING.md for guidelines.
- Fork the repo
- Create a feature branch (
git checkout -b feat/my-feature) - Run tests (
npm test) and typecheck (npm run typecheck) - Open a Pull Request
Security
Found a vulnerability? Please report it responsibly — see SECURITY.md.
License
All rights reserved. See LICENSE.
Documentation
Full documentation site: https://nirholas.github.io/pumpfun-claims-bot/
- Getting started covers install and first run.
- Examples has copy-paste snippets.
Star History
Support the project
If pumpfun-claims-bot saves you time, star it on GitHub. Stars are how other developers and AI agents find the repositories worth trusting, and they cost you one click.
Know someone who would use it? Post on X · Share on Bluesky · Share on LinkedIn · Submit to Hacker News · Share on Reddit
Built for AI agents too
Coding agents and LLM tooling can read this repo directly: AGENTS.md, llms.txt, llms-full.txt. Point an agent at https://github.com/nirholas/pumpfun-claims-bot and it has the context it needs.
More from the same author
- All repositories by nirholas: the full catalog, grouped by topic
- three.ws: the platform for 3D AI agents with Solana wallets, a skill marketplace and x402 payments
- Questions or ideas: open an issue or start a discussion
Contributors
Built by nirholas · @nichxbt · github.com/nirholas
來源:README.md,提交 24d8211
工具
0版本歷史
1- v1.0.1最新Sep 16, 2026
評分
已滿足適用檢查項的 40%
該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。
- 近 30 天內可連通25 分
請保持端點能夠回應 MCP 握手;SourceWeft 每 30 天探測一次。
- 已宣告認證方式10 分
請在註冊中心條目中宣告憑證請求標頭,或保持端點能夠回應探測,以便判定其認證方式。
- 已列出工具15 分
請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。
- 描述必需15 分
- 圖示10 分
- 文件15 分
- 授權條款10 分
請在儲存庫或註冊中心條目中宣告授權條款。