> For the complete documentation index, see [llms.txt](https://aftermaths.gitbook.io/10102-tokens/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://aftermaths.gitbook.io/10102-tokens/readme.md).

# Computing Tokens

A portfolio health analyzer for Ethereum wallets: evidence-based token scoring, verified swap recommendations, and an agent-accessible API.

**Computing Tokens** scans an Ethereum wallet, assesses every ERC-20 it holds on six evidence dimensions drawn from independent data sources, flags spam and honeypots, and, when a holding has a concrete problem, proposes a verified alternative or says hold.

**Live at:** [tokens.10102.io](https://tokens.10102.io) · **By:** [10102](https://10102.io)

## Key philosophy

Evidence over opinion, and every input named. A token's score is a function of:

* **Live data** from Etherscan, Ethplorer, CoinGecko, DEXScreener, DeFiLlama and honeypot.is, with a field left undefined whenever a source did not answer.
* **10102's own published research**, joined strictly by contract address: the [tokenomics rankings](/10102-tokens/tokenomics_value_capture.md) (one tier from T1 to T5 per token) and the deep-research reports (whether one exists for the address and its direction; the verdict text stays private).
* **A short curated dead-project list** (collapsed or migrated projects) that caps the score, and a **top-250 rank floor** for assets in the CoinGecko top 250 by market cap.

No language model touches the health score. Missing data lowers confidence, never the score, so an unknown token lands at "fair", not "critical". The full model is documented in [Intelligence Model (v2)](/10102-tokens/intelligence.md).

## What it does

* **Wallet scan**: detects every ERC-20 balance in a connected wallet via Alchemy (with CoinGecko for prices), then runs a two-phase filter to drop spam and dust before analysis.
* **Health assessment**: every surviving token gets a 0-100 health score and a separate 0-100 confidence, built from six evidence dimensions (safety, legitimacy, market, adoption, tokenomics, research), with named evidence items, concrete issue codes and a plain-English verdict.
* **Spam & honeypot detection**: generic pattern detection rather than per-token blocklists, plus symbol-collision/impersonation warnings.
* **Recommendations**: for tokens below the underperforming threshold with at least one concrete issue, proposes at most one replacement drawn from the tokenomics rankings (T1 to T3 peers in the same family), 10102 research-backed tokens, and Claude proposals resolved on CoinGecko. Every finalist is verified for Ethereum DEX liquidity, honeypot status and rank before it is shown; when nothing passes the answer is hold. See [Concepts & Scoring](/10102-tokens/concepts.md#recommendations).
* **Agent API**: the entire analysis pipeline is exposed as a bearer-authenticated HTTP API so autonomous agents can analyze wallets programmatically. See the [API Reference](/10102-tokens/api-reference/api.md).

## How it fits together

```
Wallet → Alchemy (balances) → 2-Phase Filtering → Batch Research → Assessment → Response
                                                         ↓
                     Etherscan, Ethplorer, CoinGecko, DEXScreener, DeFiLlama, honeypot.is
                                                         ↓
                                       Knowledge Base (Cloudflare KV)
```

Every scan leaves four things behind, all in Cloudflare Workers KV or the repository:

* an **immutable facts cache** per token (creation time, verification, license, proxy target, CoinGecko id, decimals), fetched once and kept for a year, so repeat lookups are faster and cheaper;
* a **nightly history** row per known token (price, market cap, volume, Ethereum DEX liquidity, latest health score) and a recomputed **calibration** file that measures forward returns by score bucket;
* a **shadow signal ledger** holding every assessment's dimension scores and confidences, at zero weight, as the feature vector calibration measures against;
* **recommendation feedback**: which recommendations were shown and which a user acted on, also at zero weight.

Scores are recomputed on every scan and never cached. Nothing in the ledger changes a score until a human promotes it with a backtest behind it. Both browser and API scans contribute equally.

## Tech stack

* **Framework:** Next.js 16 (App Router, Turbopack), React 19, TypeScript
* **Web3:** wagmi v2 + viem, RainbowKit v2 for wallet connection
* **Styling:** Tailwind CSS v4
* **Persistence:** Cloudflare Workers KV (`counters.10102.workers.dev`)
* **Data sources:** Alchemy (token discovery), Etherscan v2, Ethplorer, CoinGecko, DEXScreener, DeFiLlama, honeypot.is; plus 10102's tokenomics rankings and research reports joined by contract address
* **AI:** Anthropic Claude proposes swap candidates in the recommender (Venice as fallback); every proposal is verified against live data and no model touches the health score

## Where to go next

* New here → read [Concepts & Scoring](/10102-tokens/concepts.md) to understand health scores and verdicts, then [Intelligence Model (v2)](/10102-tokens/intelligence.md) for the evidence dimensions and the `assessment` object.
* Building an integration → jump to the [API Reference](/10102-tokens/api-reference/api.md), starting with the [overview](/10102-tokens/api-reference/api.md) for auth and rate limits.
* Reading the tokenomics tiers → see [Tokenomics rankings: methodology](/10102-tokens/tokenomics_value_capture.md) for the score, the twelve axes, the red flags and the tier gates.

{% hint style="info" %}
This is informational tooling, not financial advice. Always do your own research before acting on any score or suggestion.
{% endhint %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://aftermaths.gitbook.io/10102-tokens/readme.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.