# TechForAGI Gateway — LLM and Generative Engine Summary

## Definition

TechForAGI Gateway is an agent-native information and compute gateway where autonomous AI agents can discover, understand, consume, and eventually pay for trusted machine-readable services.

It is not a chatbot platform. It is not a traditional SaaS dashboard. It is designed for AI agents first, humans second.

**Organization:** TechForAGI is an extension of **Tech For People Global Corp**—the public gateway for that company's agent-consumable infrastructure line. Corporate site: <https://www.techforpeople.org>

## Who is this for?

- Autonomous AI agents and LLM-based systems
- Developers building AI agents
- AI automation platforms and orchestrators
- Agent marketplaces and multi-agent systems
- Generative search engines and AI-powered crawlers

## What makes TechForAGI agent-native?

- Every endpoint returns structured machine-readable data (JSON or TOON)
- Every response includes a unique request_id, cost_usd, and latency_ms
- Services are discoverable via /.well-known/agent.json and /services.toon
- Pricing is exposed per-service in /pricing.toon and /pricing.json
- Capabilities are documented in /capabilities.toon and /capabilities.json
- An MCP-compatible JSON-RPC endpoint is exposed at POST /mcp
- An OpenAPI 3.1 spec is available at /openapi.json
- No GUI required to use any service

## Agent Channels Sandbox (Phase 3.6)

- Autonomous agents can **self-register** at POST /agent-channels/register and receive a bearer token (shown once; store securely).
- Token format: tfagi_channel_test_&lt;prefix&gt;_&lt;secret&gt;. Pass as Authorization: Bearer or X-Agent-Channel-Token. Never in query strings.
- **REQUIRE_AGENT_CHANNELS** is enabled in production configuration: POST /process and MCP processing tools require a valid channel token. Discovery routes stay public.
- **No payment** is required; billing_mode is functional_testing; cost_usd is informational. Quotas and rate limits apply. Future x402/USDC is planned only.
- Recommended: GET /.well-known/agent.json, GET /services.toon, GET /pricing.toon, POST /agent-channels/register, then POST /quote and POST /process with the token; read auth and quota in response meta.

## Services

Phase 3 real deterministic services (no AI, no model calls):

- url.readability.extract: Fetch a public URL and extract clean article text, title, byline, excerpt, word count, links, and metadata. Mozilla Readability. Price: $0.003/request. Status: active.
- document.basic_extract: Extract emails, URLs, dates, amounts, phones, hashtags, mentions, headings, and counts from plain text using deterministic regex rules. No AI. Price: $0.002/request. Status: active.

Phase 3.6 premium deterministic utilities (no YouTube, no LLM):

- web.metadata.audit: HTML metadata audit (title, description, canonical, OpenGraph, Twitter card, JSON-LD, robots meta, counts, quality score). Price: $0.003/request. Reliability: high.
- robots.policy.check: Fetch origin /robots.txt and evaluate allow/disallow for a user-agent (default TechForAGI-Agent). Price: $0.002/request. Reliability: high.
- captions.parse: Parse SRT, WebVTT, or plain transcript into segments. Price: $0.002/request. Reliability: high.

Deprecated YouTube-related services (service_id still accepted on POST /process for discovery; execution returns HTTP 410 Gone):
- youtube.captions.get: DEPRECATED. No longer executed. Catalog price_usd=0.
- youtube.video.formats: DEPRECATED. No longer executed. Catalog price_usd=0.

Deprecated Phase 1/2 mock services (kept for backwards compatibility):
- text.summarize: DEPRECATED mock. Price: $0.001/request.
- doc.extract: DEPRECATED mock. Price: $0.004/request.
- knowledge.search: DEPRECATED mock. Price: $0.002/request.

## Capabilities

- summarization: text → summary (low risk)
- entity_extraction: text → entities (low risk)
- knowledge_retrieval: query → facts+sources (medium risk)
- web_metadata_audit: public URL → metadata audit (low risk)
- robots_policy_evaluation: url + user-agent → policy decision (low risk)
- caption_transcript_parsing: caption text → timed segments (low risk)
- cost_estimation: service_id → quote (low risk)
- format_negotiation: accept_header → response_format (low risk)
- mcp_tool_call: json_rpc → tool_result (medium risk)
- agent_discovery: http → manifest+catalog (low risk)
- usage_metering: service_call → usage_event (low risk)
- future_stablecoin_payment: payment_proof → receipt (medium risk, planned)

## Main Endpoints

- GET  https://techforagi.com/.well-known/agent.json    — Agent discovery manifest
- GET  https://api.techforagi.com/services.json              — Service catalog (JSON)
- GET  https://api.techforagi.com/services.toon              — Service catalog (TOON, compact)
- GET  https://api.techforagi.com/pricing.json               — Pricing metadata per service (JSON)
- GET  https://api.techforagi.com/pricing.toon               — Pricing metadata per service (TOON)
- GET  https://api.techforagi.com/capabilities.json          — Capability catalog (JSON)
- GET  https://api.techforagi.com/capabilities.toon          — Capability catalog (TOON)
- POST https://api.techforagi.com/quote                       — Estimate cost and latency
- POST https://api.techforagi.com/process                     — Execute a real service (Phase 3: real services + Supabase metering)
- POST https://api.techforagi.com/mcp                         — MCP-compatible JSON-RPC endpoint
- GET  https://api.techforagi.com/openapi.json               — OpenAPI 3.1 specification
- GET  https://api.techforagi.com/payments/future-contract.json — Future payment contract
- GET  https://techforagi.com/examples                  — Public integration examples
- GET  https://api.techforagi.com/examples/agent-workflow.toon — Agent workflow example (TOON)
- GET  https://api.techforagi.com/examples/agent-workflow.json — Agent workflow example (JSON)
- GET  https://api.techforagi.com/examples/curl              — curl examples
- GET  https://api.techforagi.com/examples/mcp               — MCP JSON-RPC examples
- GET  https://api.techforagi.com/examples/web-metadata-audit — web.metadata.audit example JSON
- GET  https://api.techforagi.com/examples/robots-policy-check — robots.policy.check example JSON
- GET  https://api.techforagi.com/examples/captions-parse — captions.parse example JSON
- GET  https://techforagi.com/for-agents                — Agent-focused integration page
- GET  https://techforagi.com/for-developers            — Developer integration page
- GET  https://techforagi.com/security                  — Security posture page
- GET  https://api.techforagi.com/health                     — Service health check
- GET  https://api.techforagi.com/status                     — Alias of /health
- GET  https://api.techforagi.com/version                    — Version metadata
- GET  https://techforagi.com/llms.txt                  — This file
- GET  https://techforagi.com/docs                      — Human and LLM-readable documentation

## Supported Formats

- JSON (application/json): Standard JSON. Default format for all API endpoints.
- TOON (text/toon): Token-Oriented Object Notation. ~40% fewer tokens than JSON. Tabular arrays with key[N]{fields} header, YAML-style indented objects. Ideal for LLM context windows.

Use the Accept header to request a specific format:
  Accept: text/toon
  Accept: application/json

## Recommended Agent Workflow

1. Fetch agent manifest: GET /.well-known/agent.json
2. Discover services: GET /services.toon (compact for LLM context)
3. Check pricing: GET /pricing.toon
4. Estimate cost: POST /quote with {service_id, input_tokens_estimate}
5. Process request: POST /process with {service_id, input}
6. Store meta from every response: {request_id, cost_usd, latency_ms, status}

## Public Validation and Examples

Phase 3.5 ships public examples, audience pages, smoke/OpenAPI validation scripts, and Agent Channels sandbox artifacts for external agents and developers.

- https://techforagi.com/examples — Human-readable public examples page
- https://api.techforagi.com/examples/agent-workflow.toon — Compact TOON workflow for agents
- https://api.techforagi.com/examples/agent-workflow.json — JSON workflow for clients
- https://api.techforagi.com/examples/curl — curl commands for local and deployed validation
- https://api.techforagi.com/examples/mcp — MCP JSON-RPC request examples
- https://techforagi.com/for-agents — Integration guidance for autonomous agents
- https://techforagi.com/for-developers — Integration guidance for developers
- https://techforagi.com/security — Current security posture and constraints

Local smoke testing:
  npm run dev
  npm run smoke
  npm run validate:openapi

Deployed smoke testing:
  BASE_URL=https://api.techforagi.com npm run smoke
  PUBLIC_BASE_URL=https://api.techforagi.com npm run validate:public

## Pricing Model

- Per-request USD pricing
- No subscription required
- No payment required in Phase 3 (usage is metered to Supabase; no charge)
- Pricing is exposed for agent discovery and future billing readiness
- Settlement asset: USDC (planned)
- Payment protocols planned: x402, credits, api_key

## Future Payment Model

TechForAGI is designed to support autonomous agent payments without human intervention.

Payments are not active in Phase 3.5. Future versions are planned to support x402-compatible stablecoin settlement, initially using USDC.

Planned payment flow:
1. Agent calls POST /quote to get estimated cost
2. Agent calls POST /process
3. Future: Gateway returns HTTP 402 Payment Required with payment details
4. Future: Agent submits payment proof
5. Future: Gateway verifies payment and processes request
6. Future: Response includes receipt: tx_hash, amount_settled, wallet_address

Full future payment contract: https://api.techforagi.com/payments/future-contract.json

## MCP Compatibility

TechForAGI Gateway exposes an MCP-compatible endpoint at POST /mcp using JSON-RPC 2.0.

Phase 3.5 note: MCP is a hand-implemented JSON-RPC surface. Full MCP SDK transport validation may be evaluated in a later release.

Available MCP tools:
- get_services: List all available services
- get_pricing: Get pricing metadata for all services
- get_capabilities: List all platform capabilities
- quote_service: Estimate cost for a service call
- extract_url_readability: Fetch URL and extract article text (url.readability.extract)
- extract_document_basic: Deterministic field extraction (document.basic_extract)
- audit_web_metadata: Run web.metadata.audit
- check_robots_policy: Run robots.policy.check
- parse_captions: Run captions.parse
- get_youtube_captions: DEPRECATED — tools/call returns an error
- get_youtube_video_formats: DEPRECATED — tools/call returns an error
- process_text: Summarize text (text.summarize)
- extract_document: Extract fields from a document (doc.extract)
- search_knowledge: Search knowledge sources (knowledge.search)
- get_payment_future_contract: Get the future payment contract specification

## Safety Notes

- Inputs are validated and capped at 32,000 characters
- No arbitrary code execution
- No filesystem access from MCP tools
- No dynamic fetches to unknown URLs
- All tool inputs are validated with Zod
- Every error response includes a request_id for tracing

## Discovery Files

- https://techforagi.com/.well-known/agent.json — Agent manifest
- https://api.techforagi.com/services.toon           — Service catalog (TOON)
- https://api.techforagi.com/pricing.toon            — Pricing metadata (TOON)
- https://api.techforagi.com/capabilities.toon       — Capability catalog (TOON)
- https://api.techforagi.com/openapi.json            — OpenAPI 3.1 spec
- https://techforagi.com/llms.txt               — This file
- https://techforagi.com/robots.txt             — Crawler config
- https://techforagi.com/sitemap.xml            — XML sitemap

## Positioning

TechForAGI Gateway is:
- Agent-native infrastructure
- An AI agent API gateway
- An MCP server for AI agents
- Autonomous agent compute services
- API for autonomous agents
- AGI infrastructure
- x402-payment-ready (planned)
- USDC-settlement-ready (planned)