Agent-native product workspace

Built for AI agents,
not just for humans.

A growing share of product work is being done by AI agents acting on behalf of PMs. Specky exposes its Product Graph, document store, ticket gen, and research tools as MCP and REST — so your agent can do PM work with the same context as the human.

Get an API key Read the spec

Interfaces

Three ways your agent talks to Specky

MCP for tool-using agents, REST for everything else, agents.json for discovery.

Beta

MCP server

Drop Specky into Claude Desktop, Cursor, or any MCP-aware agent. The server exposes searchWorkspace, createDocument, createTickets, startResearchSession, runCustomWorkflow, and the rest of the in-app toolset.

Beta

OpenAPI 3.1

Auto-generate clients in any language. Today covers capture and workspace reads; expanding to full CRUD over documents, tickets, decisions, OKRs, and experiments.

Stable

agents.json manifest

A single discovery document at /.well-known/agents.json. Crawlers and agent frameworks read it to learn Specky's capabilities, auth, and rate limits.

Planned

Outbound webhooks

Subscribe to Product Graph events — document.created, insight.created, experiment.completed — so your agent reacts instead of polls. HMAC-signed.

Quickstart

From zero to first action in 3 minutes

Generate a workspace API key

Specky → Settings → Developer → Create API key. Keys are scoped to a single workspace and follow the prefix sk_live_…

Point your agent at Specky

Add the MCP endpoint to your client config, or call /openapi.json against the REST surface. Use Authorization: Bearer <key> on every request.

Act on the user's behalf

Search the graph, draft a PRD, generate tickets, kick off research — every action lands in the PM Inbox where a human reviews before shipping.

Capture a signal

curl -X POST https://specky.space/api/extension/capture \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Competitor launched a new pricing page with usage-based tiers.",
    "url": "https://example.com/pricing",
    "source_type": "COMPETITIVE_INTEL"
  }'

Capabilities

What your agent can do

A subset of the same tools the in-app AI chat uses today — exposed as MCP tools and (progressively) as REST endpoints.

search_workspace

Hybrid ILIKE + pgvector search across every signal.

read_workspace_item

Fetch any graph node by id — doc, ticket, insight, decision, OKR.

create_document

Draft PRDs, research docs, and notes. Auto-indexed into the graph.

create_tickets

Generate engineering tickets with acceptance criteria. Push to Jira or Linear.

create_insight

Record customer insights with evidence links.

create_opportunity

Add to the Opportunity Tree under a chosen outcome.

create_experiment

Spin up A/B tests, fake doors, prototype tests.

create_decision

Log strategic decisions with rationale and citations.

create_okr

Create OKRs with key results tied to the graph.

start_research_session

Launch an Alex JTBD campaign with a shareable link.

run_custom_workflow

Execute a workspace-defined multi-step workflow.

list_inbox

Surface items awaiting human review.

Guidelines

How to be a good agent in Specky

Human-in-the-loop by default

Create items in draft / proposed state. Specky's PM Inbox is the review queue — auto-publish only when the user explicitly asks for it in the same turn.

Cite the graph

When drafting PRDs, insights, or decisions, link evidence_node_ids back to source nodes. Specky surfaces citations inline so PMs can audit your reasoning.

Treat data as confidential

Workspace data is the user's IP. Don't transmit graph contents to third-party services that aren't the user's own. Respect plan-based rate limits — back off on 429.

Ship agents that ship product.

Build on Specky's Product Graph, or plug Specky into the agent you already use.

Create API key agents.json

Agent-native product workspace

Built for AI agents,
not just for humans.

Get an API key Read the spec

Interfaces

Three ways your agent talks to Specky

MCP for tool-using agents, REST for everything else, agents.json for discovery.

Beta

MCP server

Beta

OpenAPI 3.1

Auto-generate clients in any language. Today covers capture and workspace reads; expanding to full CRUD over documents, tickets, decisions, OKRs, and experiments.

Stable

agents.json manifest

A single discovery document at /.well-known/agents.json. Crawlers and agent frameworks read it to learn Specky's capabilities, auth, and rate limits.

Planned

Outbound webhooks

Subscribe to Product Graph events — document.created, insight.created, experiment.completed — so your agent reacts instead of polls. HMAC-signed.

Quickstart

From zero to first action in 3 minutes

Generate a workspace API key

Specky → Settings → Developer → Create API key. Keys are scoped to a single workspace and follow the prefix sk_live_…

Point your agent at Specky

Add the MCP endpoint to your client config, or call /openapi.json against the REST surface. Use Authorization: Bearer <key> on every request.

Act on the user's behalf

Search the graph, draft a PRD, generate tickets, kick off research — every action lands in the PM Inbox where a human reviews before shipping.

Capture a signal

curl -X POST https://specky.space/api/extension/capture \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Competitor launched a new pricing page with usage-based tiers.",
    "url": "https://example.com/pricing",
    "source_type": "COMPETITIVE_INTEL"
  }'

Capabilities

What your agent can do

A subset of the same tools the in-app AI chat uses today — exposed as MCP tools and (progressively) as REST endpoints.

search_workspace

Hybrid ILIKE + pgvector search across every signal.

read_workspace_item

Fetch any graph node by id — doc, ticket, insight, decision, OKR.

create_document

Draft PRDs, research docs, and notes. Auto-indexed into the graph.

create_tickets

Generate engineering tickets with acceptance criteria. Push to Jira or Linear.

create_insight

Record customer insights with evidence links.

create_opportunity

Add to the Opportunity Tree under a chosen outcome.

create_experiment

Spin up A/B tests, fake doors, prototype tests.

create_decision

Log strategic decisions with rationale and citations.

create_okr

Create OKRs with key results tied to the graph.

start_research_session

Launch an Alex JTBD campaign with a shareable link.

run_custom_workflow

Execute a workspace-defined multi-step workflow.

list_inbox

Surface items awaiting human review.

Guidelines

How to be a good agent in Specky

Human-in-the-loop by default

Create items in draft / proposed state. Specky's PM Inbox is the review queue — auto-publish only when the user explicitly asks for it in the same turn.

Cite the graph

When drafting PRDs, insights, or decisions, link evidence_node_ids back to source nodes. Specky surfaces citations inline so PMs can audit your reasoning.

Treat data as confidential

Workspace data is the user's IP. Don't transmit graph contents to third-party services that aren't the user's own. Respect plan-based rate limits — back off on 429.

Ship agents that ship product.

Build on Specky's Product Graph, or plug Specky into the agent you already use.

Create API key agents.json

Built for AI agents, not just for humans.

Three ways your agent talks to Specky

MCP server

OpenAPI 3.1

agents.json manifest

Outbound webhooks

From zero to first action in 3 minutes

Generate a workspace API key

Point your agent at Specky

Act on the user's behalf

What your agent can do

How to be a good agent in Specky

Human-in-the-loop by default

Cite the graph

Treat data as confidential

Ship agents that ship product.

Built for AI agents, not just for humans.

Three ways your agent talks to Specky

MCP server

OpenAPI 3.1

agents.json manifest

Outbound webhooks

From zero to first action in 3 minutes

Generate a workspace API key

Point your agent at Specky

Act on the user's behalf

What your agent can do

How to be a good agent in Specky

Human-in-the-loop by default

Cite the graph

Treat data as confidential

Ship agents that ship product.

Built for AI agents,
not just for humans.

Built for AI agents,
not just for humans.