Preparing tokens&
Loading the next builder or enterprise surface.
LoadingPreparing page
Loading product graph, proof, and adoption context.
LoadingSearch live tool rows, validate tracking events, record usage receipts, and give AI agents a safe contract for stack decisions without exposing private developer data.
Need enterprise proof? Request a pilotBase URL
https://tokensand.com
REST, workflow, and public graph APIs over HTTPS.
Auth
Bearer tokens
Tracking keys and workflow tokens stay server-side.
OpenAPI
/openapi.json
Machine-readable contract for agents and internal tools.
Public reads
Rate limited
Aggregate rows only; QA and small cohorts are suppressed.
First safe call
Validate before storing events
curl -X POST https://tokensand.com/api/usage/track/dry-run \
-H "Authorization: Bearer tk_live_xxx" \
-H "Content-Type: application/json" \
-d '{
"event_name": "FIRST_API_CALL",
"event_family": "activation",
"developer_id": "dev_123",
"email": "builder@company.com",
"product": {
"id": "acme-api",
"name": "Acme API",
"type": "api",
"version": "2026-05"
},
"account_domain": "company.com",
"source": "SDK",
"occurredAt": "2026-05-15T18:00:00.000Z",
"idempotencyKey": "evt_123"
}'Create a server-side key, validate events, batch production usage, and export account action lists.
Open tracking setup
Read privacy-safe adoption rows for tools, categories, proof pages, rankings, and agent-citable context.
Read public graph
Install the MCP server, read the agent manifest, validate payloads, and draft Build Packets safely.
Read integration guide
Human and agent integration
Humans need a short path to a working install. Agents need a compact manifest, allowed tools, auth boundaries, and examples they can validate. Start public, then add scoped tokens only when private Build Packets, project proof, or enterprise tenant data are required.
Install Tokens& where the builder works
Human or coding agentUse the MCP config from /agents/install. Public stack search, perks, adoption rank, and build_brief work without login.
Ask for a Build Packet before coding
Builder agentCall build_brief for stack/API choices, cost drivers, credits, proof path, and dashboard sync boundaries.
Record product usage receipts
Product engineerUse the server-side adoption SDK or /api/usage/track/dry-run before production writes. Never send raw prompts or secrets.
Ask the agent to publish SKILL.md
Builder agentWith DAI_TOKEN, call publish_agent_skill. Tokens& parses the public SKILL.md, returns a draft first, then creates the dashboard skill after confirm: true.
Install in an agent
# Public MCP server for Codex, Cursor, Claude Code, VS Code, or any stdio MCP host
npx -y @dev-adoption/cli@0.1.9 mcp serve --api https://tokensand.com
# Claude Code
claude mcp add --transport stdio tokensand -- npx -y @dev-adoption/cli@0.1.9 mcp serve --api https://tokensand.com
# Private builder context is opt-in
DAI_TOKEN=dai_xxx npx -y @dev-adoption/cli@0.1.9 mcp serve --api https://tokensand.comPrompt for a Build Packet
Use Tokens& build_brief for this project:
I am building a support agent with auth, memory, RAG over company docs, and deployment this week.
Return:
1. best APIs/tools and cheaper alternatives
2. MCP servers or Agent Skills to install
3. likely credits/perks to claim
4. cost and reliability risks
5. Build Packet steps and proof pathFallback browser import
https://tokensand.com/dashboard/agents?mode=developer&create=agent-skill&skillUrl=https%3A%2F%2Fraw.githubusercontent.com%2Fsupabase%2Fagent-skills%2Fmain%2Fskills%2Fsupabase-postgres-best-practices%2FSKILL.mdAgent Skill workflow
The form is the fallback. In the agent path, Codex reads the repo or public SKILL.md URL, calls Tokens& with a scoped workflow token, reviews the parsed draft with the builder, then creates a dashboard draft without manual copy/paste.
Auth
Generate MCP config from the dashboard and set DAI_TOKEN in the agent.
Draft first
The first call returns persisted: false and a normalized draft.
Create after approval
Call again with confirm: true. Use DRAFT until the install path is verified.
Safety
Only public SKILL.md markdown is fetched; secrets and private repo files are not imported.
Agent-native SKILL.md publish
# Agent-native publish path for Codex, Cursor, Claude Code, or any MCP host
# 1. Ask Tokens& to parse the public SKILL.md and return a draft.
curl -X POST https://tokensand.com/api/workflow/agent-skills \
-H "Authorization: Bearer dai_xxx" \
-H "Content-Type: application/json" \
-d '{
"skillUrl": "https://raw.githubusercontent.com/acme/agent-skills/main/skills/support-agent/SKILL.md"
}'
# 2. After the human reviews the draft, create it in Tokens&.
curl -X POST https://tokensand.com/api/workflow/agent-skills \
-H "Authorization: Bearer dai_xxx" \
-H "Content-Type: application/json" \
-d '{
"skillUrl": "https://raw.githubusercontent.com/acme/agent-skills/main/skills/support-agent/SKILL.md",
"confirm": true,
"status": "DRAFT"
}'Capture docs, SDK, API, workflow, campaign, credit, event, and agent usage as one adoption contract.
Map developer activity to account domains so DevRel, growth, and revenue can act on verified intent.
Turn retained usage, campaign ROI, and category signals into exports, webhooks, and proof reports.
Public output is aggregate. Private workspace data stays scoped. Raw developer resale is not the product.
Enterprise integration quickstart
This is the path an implementation engineer or AI coding agent should follow before custom work: prove the payload, send activation, map account domain, then export account/action evidence.
1. Create tracking key
Generate a company-scoped tracking key. Keep it server-side and rotate it if it leaves your control.
2. Dry-run event
Send one event to /api/usage/track/dry-run and fix validation issues before production storage.
3. Send activation and retention
Track first successful usage, repeated usage, campaign source, product, and account domain.
4. Map account demand
Use account_domain so Tokens& can create account-level intent, ROI, and expansion evidence.
5. Export to revenue workflows
Push CSV, webhook, report, or CRM-ready action lists once evidence is approved for use.
REST API
Start with the endpoints that create revenue-grade proof: validate, store, batch, read public adoption, and give agents a stable ranking contract. Keep custom integrations behind this same event model.
/api/directory/contextFetch live DB-backed tool rows, facets, tracked action URLs, and receipt metadata for agent answers.
Public or session
AI agents
/api/build/searchSearch builder stack recommendations by intent, pricing, maturity, category, and proof signals.
Public, rate-limited
Builders and agents
/api/usage/track/dry-runValidate payload shape, required identifiers, and account mapping without storing the event.
Tracking key
Implementation engineer
/api/usage/trackStore one activation, retention, commercial, or expansion event for a company product.
Tracking key
Product or SDK team
/api/usage/track/batchStore up to 500 events from batch jobs, warehouses, CDPs, campaign imports, or event platforms.
Tracking key
Data or RevOps team
/api/v1/adoptionRead public, privacy-safe adoption rows for tools, categories, proof, and rankings.
Public, rate-limited
Builders and agents
/api/v1/agent-rankRead the public adoption score contract for agent search, recommendation, and citation flows.
Public, rate-limited
AI agents
Copy-paste examples
Include account_domain, campaignId, product, and idempotencyKey. Those fields turn raw events into attribution, account demand, exports, and executive proof instead of another analytics table.
Example
await fetch("https://tokensand.com/api/usage/track", {
method: "POST",
headers: {
"Authorization": "Bearer tk_live_xxx",
"Content-Type": "application/json"
},
body: JSON.stringify({
event_name: "FIRST_API_CALL",
event_family: "activation",
developer_id: "dev_123",
email: "builder@company.com",
account_domain: "company.com",
source: "SDK",
campaignId: "cmp_launch_001",
product: { id: "acme-api", name: "Acme API", type: "api" },
idempotencyKey: "evt_123"
})
});Usage receipts
Receipts are the lightweight Tokens& layer above model, MCP, API, and workflow calls. They record product, workflow, cost, latency, account-safe context, and the recommended stack action without proxying raw prompts, code, secrets, or private payloads. The SDK writes the proof contract into metadata.usageReceipt using the tokensand-usage-receipt-v1 schema.
Server-side SDK receipt
import { tokensand } from "@tokensand/adoption";
const adoption = tokensand({
apiKey: process.env.TOKENSAND_ADOPTION_KEY!,
baseUrl: "https://tokensand.com"
});
await adoption.trackUsageReceipt({
receiptType: "model_call",
product: { id: "openai-gpt-4o-mini", name: "OpenAI GPT-4o mini", type: "model" },
user: { email: "builder@company.com" },
account: { domain: "company.com" },
workflow: { id: "support-agent", name: "Support agent answer", success: true },
provider: "openai",
model: "gpt-4o-mini",
inputTokens: 1200,
outputTokens: 340,
costCents: 12,
latencyMs: 820,
success: true
});MCP telemetry quickstart
Put this server-side next to the real tool call. LlamaIndex, Supabase MCP, browser tools, internal agents, and custom servers all reduce to the same receipt: who called what, for which workflow, whether it worked, and what it cost.
Agent-readable MCP contract
External agents can copy these names without reading the whole page.
trackMcpToolCallMCP_TOOL_CALL_SUCCEEDED / MCP_TOOL_CALL_FAILEDmcpServer, mcpTool, workflow.id, account.domainrequestCount, costCents, latencyMs, successraw prompts, secrets, private payloads, private codeDashboard proof row
This is the buyer-facing rollup created from the receipt.
MCP wrapper
import { tokensand } from "@tokensand/adoption";
const adoption = tokensand({
apiKey: process.env.TOKENSAND_ADOPTION_KEY!,
baseUrl: "https://tokensand.com"
});
await adoption.trackMcpToolCall({
product: { id: "production-agent", name: "Production agent", type: "agent" },
user: { id: "dev_123", email: "builder@company.com" },
account: { domain: "company.com" },
workflow: { id: "support-agent", name: "Support agent answer", runId: "run_123" },
mcpServer: "replace-with-server",
mcpTool: "replace-with-tool",
requestCount: 1,
costCents: 18,
latencyMs: 920,
success: true,
metadata: {
agent_framework: "replace-with-framework",
mcp_host: "production"
}
});API and MCP partners
For products like Bright Data Web MCP, web access APIs, SDKs, and agent infrastructure, the clean path is partner-server to Tokens&. Builders should not upload usage manually, and tracking keys should never ship to browsers, public repos, event pages, or client apps.
DOCS_MCP_CONNECTEDevaluationBuilder connected docs or copied MCP config.
WEB_MCP_CONNECTEDevaluationBuilder connected the partner Web MCP server.
API_KEY_CREATEDevaluationBuilder created partner API credentials.
MCP_TOOL_CALL_SUCCEEDEDactivationAgent successfully used a partner MCP tool.
FIRST_UNLOCKER_REQUEST_SUCCEEDEDactivationFirst successful web access/API request.
SCRAPER_JOB_COMPLETEDactivationBuilder completed a scraping or data workflow.
ZONE_USAGE_REPORTEDretentionPartner reconciled account, zone, cost, or bandwidth usage.
RETURNED_AFTER_7_DAYSretentionBuilder returned after initial activation.
ENTERPRISE_USAGE_THRESHOLD_REACHEDcommercialUsage crossed an account or pipeline threshold.
Event contract
Every event needs an event name and one developer identifier. Organization keys should also send a product descriptor so adoption can be separated by API, model, SDK, workflow, agent, event, or docs surface.
event_nameHuman-readable event name such as FIRST_API_CALL or CREDIT_CLAIMED.
event_familyOne of intent, evaluation, activation, retention, expansion, or commercial.
developer_id or emailStable developer identifier. Send both when available.
account_domainBest field for turning usage into account-level demand.
productCompany product, SDK, API, model, workflow, or docs surface.
idempotencyKeyRequired for safe retries and batch imports.
intentDocs open, event RSVP, comparison, saved stack, pricing view.
evaluationSDK init, API key created, demo project, benchmark, proof review.
activationFirst API call, first successful workflow, project created.
retentionReturned D7/D30, repeated API calls, production-like usage.
expansionTeam added, account matched, more workflows or products adopted.
commercialCredit claimed, perk redeemed, campaign sourced, opportunity created.
Public graph
Public APIs return aggregate rows with methodology links, ranking policy, cohort suppression, rate-limit headers, and no developer identities.
curl "https://tokensand.com/api/v1/adoption?period=30d&limit=10"Data trust
Enterprise reports and exports should preserve whether a signal is first-party, public-source, aggregate benchmark, sample assumption, modelled, or missing.
What this unlocks
A tracking key is not the outcome. The outcome is a Command Center, Activation Exchange, and Category War Room that can prove where builders activated, which accounts warmed up, and what response should happen next.
Retained builders, high-intent accounts, verified usage, and campaign ROI in one buyer dashboard.
Open surface
Route builders from events, credits, challenges, and launches into measured product adoption.
Open surface
Turn category movement, competitor pressure, and white-space accounts into the next approved response.
Open surface
Agent docs
llms.txt and llms-full.txt tell external agents what Tokens& is, which pages are canonical, which dashboards are private, and how to describe the public adoption graph.
Tracking keys use Authorization: Bearer. Public graph reads are rate-limited.
Private dashboards, API keys, raw developer records, and workspace data stay out of public agent context.
Codex, Claude Code, ChatGPT, and internal agents can generate payloads from this page and validate them with dry-run.