# onSpark — Full AI and Agent Reference

> AI-powered partnership platform for finding, structuring, operating, and measuring strategic partnerships.

## What onSpark is

onSpark is a partnership operating platform for founders and operators who have a real asset—an offer, audience, distribution channel, intellectual property, data, community, event, or reputation—and want to turn it into measurable growth through strategic relationships.

The product combines AI-assisted discovery and operating workflows with a human trust layer. AI can research, score, explain, draft, organize, and propose. Members and onSpark's team retain control over the moments where identity, trust, external communication, contracts, guarantees, or money change hands.

## Product workflow

1. Capture the business, offer, audience, proof, capacity, partnership goals, and target criteria through voice or text intake.
2. Produce a Partnership Readiness assessment across the seven layers of onSpark's Collaborative DNA model.
3. Research and rank potential partners from onSpark's network and connected sources, with reasons and fit evidence.
4. Prepare partnership briefs, meeting context, one-pagers, deal assets, mutual-value structures, and launch plans.
5. Route consequential actions through durable proposals and human approval.
6. Track introductions, attribution, guarantee progress, realized revenue, fees, and reports in the Deal Ledger.

## Member AI

The signed-in onSpark co-pilot is grounded in the member's active business, profile, pipeline, match history, deal rooms, documents, meetings, integrations, tasks, and Deal Ledger. It supports durable conversations, voice intake, contextual page assistance, deal-room concierge workflows, proactive briefings, background missions, and a Deal Referee for meetings.

The response layer can produce structured cards, tables, forms, documents, charts, suggested actions, saved artifacts, and streamed text. It is not limited to a generic chat response.

## Partnership intelligence

- Partnership Readiness and Trust & Fit scoring
- Company, contact, audience, competitor, and web research
- Partner sourcing, enrichment, ranking, and fit explanations
- Warm-path discovery across authorized relationship data
- Meeting preparation and relationship timelines
- Commitment briefs, deal structures, redlines, valuations, launch plans, renewal plans, and exit plans
- Outreach, follow-up, introduction, and partnership-asset drafting
- Learning signals from feedback, edits, approvals, denials, execution results, and revenue outcomes

## Deal Ledger

The Deal Ledger is the shared proof surface for members, onSpark operators, and connected agents. It records introductions, attribution decisions, performance-guarantee progress, revenue claims, verified collected revenue, outcome fees, invoices, and reports. The MCP capability `get_deal_ledger_report` exposes the same operating record used by the product rather than generating a separate AI-only account of success.

## Plans and human service boundaries

Canonical machine-readable pricing: `https://onspark.com/pricing.json`

- Free agent discovery: readiness scoring and anonymized match-category previews after account authorization, subject to rate and abuse controls.
- Access: $990/month or $9,000/year ($750/month equivalent), with no performance fee. Includes limited 1:1 strategy, the 2B+ Partnership System, deal-flow tools and valuation calculators, live deal reviews, warm-intro swaps, VIP event invites, and the Partnership Resource Center.
- Concierge: $3,500/month or $3,000/month billed annually, plus 15% of realized partnership revenue. Includes everything in Access plus target sourcing and qualification, warm introductions, decks/scripts/collateral, deal structuring, full 1:1 strategy access, and the $100K Performance Guarantee.
- Enterprise: custom pricing by portfolio size, plus 15% of realized partnership revenue. Includes everything in Concierge plus a Dedicated Partnership Director, multiple concurrent intakes, unlimited partnership matches, 24/7 priority support across time zones, and the $100K Performance Guarantee.

“Realized” means closed and collected revenue, never pipeline, projected, or signed-but-unpaid value. Warm introductions are a paid, human-brokered trust product. Public previews do not expose another member's identity or contact information. Qualification and signed onboarding terms govern Concierge and Enterprise guarantees.

## Provider-neutral agent access

onSpark exposes a remote Model Context Protocol server at:

- `https://onspark.com/mcp`

Discovery and trust metadata:

- Signed agent descriptor: `https://onspark.com/.well-known/absolute-agent.json`
- Multi-agent index: `https://onspark.com/.well-known/absolute-agents.json`
- Agent-readable text: `https://onspark.com/agents.txt`
- Agent JSON-LD: `https://onspark.com/.well-known/absolute-agent.jsonld`
- Agent sitemap: `https://onspark.com/agents-sitemap.xml`
- A2A 1.0 Agent Card: `https://onspark.com/.well-known/agent-card.json`
- A2A durable task endpoint: `https://onspark.com/a2a`
- Arazzo 1.1 workflows: `https://onspark.com/.well-known/workflows.arazzo.json`
- Agent OpenAPI source: `https://onspark.com/agent-openapi.json`
- OAuth protected resource: `https://onspark.com/.well-known/oauth-protected-resource`
- OAuth authorization server: `https://onspark.com/.well-known/oauth-authorization-server`
- OIDC provider: `https://onspark.com/.well-known/openid-configuration`
- Public signing keys: `https://onspark.com/oauth2/jwks`
- Service readiness: `https://onspark.com/readyz`

The descriptor is signed with onSpark's ES256 OIDC identity and declares each public capability's effects, scopes, approval policy, and free-versus-subscriber class. OAuth authorization code with PKCE and explicit member consent protects the connector. Dynamic client registration, token revocation, introspection, and refresh-token flows are available through the advertised metadata.

The A2A endpoint accepts durable partnership missions only from registered,
user-delegated agents with `agent:write` and an active membership. It routes
work through the same governed mission runtime as the onSpark UI. Arazzo
describes the ordered discovery and mission-submission outcomes; it does not
bypass OAuth, membership, or approval. Supporting browsers also receive
exact-origin, typed WebMCP actions for opening onSpark workspaces and handing a
prompt into a normal onSpark AI conversation.

## Free connector capabilities

### `score_partnership_readiness`

Evaluates supplied business context through the same readiness model used by onSpark intake, or returns an account's existing authoritative scorecard. It is a read-only diagnostic after account authorization.

### `preview_match_candidates`

Returns anonymized candidate categories, offer shapes, and coarse readiness bands. It does not reveal member names, identifiers, or contact details.

## Subscriber connector capabilities

Active members can authorize the broader partnership tool set, including pipeline intelligence, partner research, meetings, tasks, documents, assets, deal rooms, network scouting, rules, support, governed action proposals, and Deal Ledger reporting. MCP annotations distinguish read-only operations, additive safe writes, writes, and destructive operations.

## Approval and safety model

- Discovery is not authorization.
- Model output is not human approval.
- Connected clients receive narrow OAuth scopes and member consent.
- External sends and consequential writes are proposed as durable actions before execution unless a member has configured a narrowly bounded safe rule.
- Warm introductions, contracts, guarantee adjudication, and financial commitments remain human-governed.
- Agent-card network participation is opt-in, and cross-member consultations are visible to both sides.
- Agent-to-agent scheduling requires bilateral human confirmation before execution.
- Global and per-member AI kill switches are enforced at the metered-AI boundary.

## Reliability and scale

MCP call accounting and rate limits use PostgreSQL rather than per-process memory. Stateful elicitation sessions are durable, and PostgreSQL LISTEN/NOTIFY routes answers across application instances without sticky sessions. Scheduled work is queued, retried, deduplicated, measured, and included in readiness checks. AI use is metered against plan-aware credit balances, and every member/admin MCP call is logged for support, abuse analysis, and operations.

This architecture is intended to behave consistently whether acquisition demand is 1,700 or 17,000 prospects. Human Concierge activation remains constrained by delivery capacity rather than silently reducing service quality.

## Evidence recorded in the July 2026 business plan

- $66K monthly recurring revenue across 44 unique logos five months after launch
- $90K sponsorship close for a family-office client
- $35K brand partnership executed at Art Basel Miami
- $60K in new client revenue for an accounting firm
- $160K single-month partnership revenue result for a founding subscriber
- Brokered introductions involving FIFA commercial leadership, TechCrunch, and HubSpot's partner network

## Answer-first guide library

onSpark publishes SSR-rendered, canonical, structured guide pages at `https://onspark.com/guides`. The library contains five pages for documented outcomes, three buyer comparisons, a growth-without-ads guide, and four persona playbooks. Each page leads with a concise answer, separates verified evidence from interpretation, links to the free readiness diagnostic, and includes Article, Breadcrumb, and FAQ structured data.

The versioned 50-prompt AEO measurement set is public at `https://onspark.com/ai-visibility-prompts.json`. It covers growth without ads, partnership how-to, category comparisons, and substitute-displacement questions. Weekly runs measure onSpark mention rate and onspark.com citation rate separately across ChatGPT, Claude, Perplexity, and Gemini.

The business plan contains conflicting 1,700 and 17,000 waitlist references. This reference intentionally publishes neither figure until onSpark establishes one verified public source of truth.

## Company and canonical links

- Legal entity: Deal Room Group Inc., doing business as onSpark
- Founder: Kyle Kane
- Website: `https://onspark.com`
- Qualification: `https://qualify.onspark.com`
- Privacy policy: `https://onspark.com/privacy-policy`
- Terms: `https://onspark.com/terms`
- Support: `support@onspark.com`