GraphIQ.ai changelog
Canonical page: https://graphiq.ai/resources/changelog/
RSS: https://graphiq.ai/resources/changelog/feed.xml
Last updated: 2026-10-14
Every GraphIQ.ai platform, API and MCP release, newest first. Each release is tagged with the layer of the graph it strengthens:
Structure: How companies are organized: parents, subsidiaries, locations and corporate families.
Facts: Current, verified facts attached to exactly the right company or person.
Connections: The relationships that let you move from one company to the next like it.
Access: How people, AI agents and models query the graph and act on it.
MCP 2.0: four times the reach for your AI agent (October 2026)
Released: 2026-10-14
Layer: 4 · Access (Locations Ontology: 1 · Structure)
Tags: MCP, Locations Ontology, Web app
For: Agent builders, RevOps, Sales
Where: MCP · Product UI
Permalink: https://graphiq.ai/resources/changelog/#oct-2026 (also #mcp-2-0)
Summary. Your AI agent can now do real go-to-market research. With MCP 2.0, any MCP-compatible AI tool connects straight to the GraphIQ Identity Graph. That includes 287 million organizations, 393 million people and more than 2 billion news articles. Just ask in plain English. You get account research, market maps, prospect lists and news alerts back, with no exports and no copy-paste.
17 agent tools (up from 4) · 6 built-in playbooks · 287M organizations · 393M people · 2B+ news articles
What shipped
MCP 2.0: 17 agent tools (up from 4), a built-in web crawl and 6 step-by-step agent playbooks
Locations Ontology: one canonical name for every place across the Identity Graph
Web app: mobile and tablet support, a new Account Settings area and People Export fixes
Why it matters. Your agent can move from "who makes this?" to "who do I call there?" to "what changed this week?" in a single conversation.
A complete toolkit for agents. MCP 2.0 adds 17 read-only tools across six areas:
Organizations: search, full company profiles and market counts
People: contacts filtered by seniority and function
News: deduplicated coverage of a company, with sentiment scores
Ontologies: capabilities, technologies, NAICS codes and places
Web: live web search and page reading
Guides: step-by-step playbooks for complex tasks
All 17 tools are read-only, so an agent can explore freely without changing your data. You don’t need to worry about the specific tool names as the AI agent is really good at figuring out which tools to use for your prompt.
Built-in web crawl that saves tokens. Agents can check any company's website without leaving GraphIQ. Pages come back clean and compact, so they use far fewer tokens than general-purpose browsing. Redirects are flagged too, so agents can spot rebrands and acquisitions.
Playbooks that make agents get it right. Even strong models make predictable mistakes with B2B data. They guess at industry names, filter out real companies because headcounts are stored as size bands, or count the same company five times because it has five branch offices. Our guides are playbooks the agent reads before it starts. Each one gives the tool sequence, a worked example and the known traps.
Company research: a grounded pre-call brief covering firmographics, organization relationships, the last 90 days of material news, key executives and a website cross-check.
Market mapping: Create a detailed market research report, the size and shape of a segment (how many companies, where they are, what sizes) before you drill in.
Prospect lists: a clean territory or account list ranked by size, with branch offices merged into one company, fragments dropped, and every surviving company checked against its own website.
Find contacts: named people with work emails at target companies, filtered by seniority and function.
News monitoring: funding, M&A, launches and leadership changes for a set of companies, without duplicates.
Query syntax: a reference for AND/OR/NOT logic, controlled vocabularies, ranges, dates and data caveats.
Get started in three steps. GraphIQ works with Claude, ChatGPT and other AI tools that support remote MCP servers. If you have a paid MCP account, connect to our MCP at https://app.graphiq.ai/mcp
Add GraphIQ to your AI tool. In your tool's connector or MCP settings, add GraphIQ as a remote MCP server using https://app.graphiq.ai/mcp.
Sign in with your GraphIQ account.
Ask a question. For multi-step work, it helps to name the playbook, for example: "Use the prospect lists guide to build me a territory list of…"
Try it (MCP endpoint)
```
https://app.graphiq.ai/mcp

# works with Claude, ChatGPT, Microsoft Copilot Studio, Cursor
# or any AI client that supports remote MCP servers
```
Three ways to connect. Remember: You can connect to GraphIQ.ai three ways - through our Ux, through our MCP, or through our API for programmatic access.
Web via Login: https://app.graphiq.ai/account
MCP via OAuth - used for chat-based agents (Claude Chat/Cowork, ChatGPT, etc.)
REST API (Bearer Authentication). Used for coding tools or systematic access (Claude Code, Codex, etc)
A single source of data for location. Our new Locations Ontology gives every country, state, county and city one canonical name across the Identity Graph. Common aliases resolve automatically, and named regions like EMEA and APAC work as filters. You can search by headquarters or by any office location. The result is more complete territory lists and more consistent geographic analysis, whether you search in the app or through an agent.
A better app experience
Mobile and tablet: GraphIQ now works well in phone and tablet browsers, so you can research accounts between meetings.
Account Settings: A new settings area under your avatar (the gear icon) brings your Profile and your Privacy & Security settings together in one place.
More reliable exports: We fixed bugs in People Export.
Coming next
News Subscriptions: Turn any saved news search into an email alert, delivered hourly, daily or weekly.
MCP 2.1: More playbooks, plus richer tool definitions and new capabilities.
Cleaner company records: Canonical Org pipeline upgrades will reduce duplicate and fragmented organization records.
Better company descriptions: An improved summary pipeline will give more accurate, informative descriptions of what each company does.
Links
Connect GraphIQ to your AI tool
Book a demo
MCP
MCP Server
Funding Intelligence & Capability Explorer Enhancements (v1.20)
Released: 2026-02-03
Layer: 2 · Facts
Tags: Signals, Entity pages
For: RevOps, Sales
Where: Product UI · API · MCP
Permalink: https://graphiq.ai/resources/changelog/#v1-20 (also #funding-intelligence)
Summary. GraphIQ.ai detects Series A–D funding rounds within 24 hours of filing and shows them in a Recent funding module on each entity page (v1.20, February 3, 2026).
What shipped
Funding round detection now surfaces Series A–D events within 24 hours of filing.
Capability Explorer updated with filterable tags across 40+ capability categories.
Entity pages now include a "Recent funding" module linked to the signals feed.
Why it matters. Revenue teams can trigger outbound on the day a prospect raises — not a week later.
The problem. Funding is one of the clearest buying signals, but it usually reaches reps late, and keyword-based feeds often attach it to the wrong company.
How it works
GraphIQ.ai watches filings and news for Series A–D rounds.
Each round resolves to one canonical entity, so the signal lands on the right company and its corporate family.
The round appears in the entity’s Recent funding module and in the Signals feed within 24 hours of filing.
Signals also reach you by email digest, webhook, API pull or the MCP server.
Example. A Series B filing for a company in your ICP shows up on its entity page the next morning, with its parent company and any other active signals alongside it.
Try it (REST API)
```
POST https://api.graphiq.ai/api/v2/news/search
X-API-Key: YOUR_WORKSPACE_KEY

# filter by organization ID, keyword, date or sentiment
```
Related
The Salesman’s Conundrum (Blog · 4 min)
AI = Models + Data: Why the Data Layer Decides Everything (Blog · 7 min)
Signals
Identity Graph
Fingerprint Search is generally available (v1.19)
Released: 2026-01-15
Layer: 3 · Connections
Tags: Search, GA release
For: Sales, RevOps
Where: Product UI
Permalink: https://graphiq.ai/resources/changelog/#v1-19 (also #fingerprint-search)
Summary. GraphIQ.ai Fingerprint Search, generally available since v1.19 (January 15, 2026), turns a plain-English market description into a capability fingerprint and searches 287 million organizations against it.
What shipped
Describe a market in plain English; the graph parses it into a capability fingerprint across 287M organizations.
Results return with entity confidence scores, hierarchy context, and signal overlays.
Why it matters. No more SIC codes or boolean strings — teams search for intent, not filters.
The problem. Traditional market search depends on SIC codes and boolean strings. Those describe how a company was classified years ago, not what it does today.
How it works
State your intent in plain English.
The graph parses it into a fingerprint across capability, scale, geography, hierarchy and signal dimensions.
Vector search and graph traversal run together across the graph.
Results come back ranked, with entity confidence scores, hierarchy context and signal overlays.
Matches are deliberately capped at a couple of thousand highly resolved entities, because accuracy drops sharply past that point.
Example query. Tier-2 precision machining suppliers to automotive OEMs in the Great Lakes region with ISO 9001 certification and 100–800 employees.
Related
The Salesman’s Conundrum (Blog · 4 min)
Knowledge Graphs for GTM (Guide · 14 min)
Fingerprint Search
Graph mutations: write-back over MCP (v1.18)
Released: 2025-12-11
Layer: 4 · Access
Tags: MCP, Governance
For: Agent builders, Developers
Where: MCP
Permalink: https://graphiq.ai/resources/changelog/#v1-18 (also #mcp-write-back)
Summary. Since v1.18 (December 11, 2025), GraphIQ.ai lets AI agents write back to the graph over MCP, flagging entities, updating watch lists and routing signals, under read/write governance set per workspace.
What shipped
Agents can flag entities, update watch lists and route signals over the MCP server.
Read/write governance is configurable per workspace.
Why it matters. Agents can act on what they find in the graph instead of only reading it, and admins decide exactly what they are allowed to change.
The problem. Most agents can read business data but can’t record what they learn, so their findings stay in a chat transcript instead of the system of record.
How it works
Connect to the MCP server over streamable HTTP. Tools are discovered on the first handshake.
Authenticate with an X-API-Key for backend and CLI use, or OAuth 2.0 for interactive web apps.
Set read/write governance for the workspace.
Agents flag entities, update watch lists and route signals as ordinary tool calls.
Every session is isolated to its own tenant.
Try it (MCP endpoint)
```
https://app.graphiq.ai/mcp

# works with Claude, OpenAI Agents SDK, LangChain, LangGraph or custom Python
```
Related
Knowledge Graphs for GTM (Guide · 14 min)
The Architecture Answer: Headless Is the Precondition, Not the Plumbing (Blog)
MCP
MCP Server
Continuous contact refresh & phone verification (v1.17)
Released: 2025-11-20
Layer: 2 · Facts
Tags: People Search
For: Sales
Where: Product UI · API
Permalink: https://graphiq.ai/resources/changelog/#v1-17 (also #contact-refresh)
Summary. Since v1.17 (November 20, 2025), GraphIQ.ai refreshes its 393 million employee records continuously and marks every direct-dial number with a freshness score and last-verified timestamp.
What shipped
393M employee records now refresh continuously rather than on a quarterly batch schedule.
Direct-dial numbers carry a freshness score and last-verified timestamp.
Why it matters. Reps stop dialing disconnected numbers and emailing people who changed jobs two quarters ago.
The problem. About 30% of a B2B contact list goes out of date each year. With quarterly batch refreshes, titles and numbers are already wrong on the day you export them.
How it works
Employee records re-resolve as new information arrives, instead of waiting for a quarterly batch.
Each direct dial carries a freshness score and a last-verified timestamp, so you can see how recently it was confirmed.
Tip. Sort call lists by freshness score so reps start with the numbers most likely to connect.
Related
B2B Relationship Intelligence (Guide · 12 min)
Context Over Quantity: Stop Doing V-Lookups at Scale (Blog · 11 min)
Lookalike discovery v2: graph-native matching (v1.16)
Released: 2025-10-30
Layer: 3 · Connections
Tags: Lookalike
For: RevOps, Developers
Where: Product UI · API · MCP
Permalink: https://graphiq.ai/resources/changelog/#v1-16 (also #lookalike-v2)
Summary. GraphIQ.ai Lookalike v2 (v1.16, October 30, 2025) matches companies on a live graph signature of funding activity, tech events and M&A, and runs programmatically with similar=True over the API and MCP server.
What shipped
Matching now runs against a dynamic graph signature — funding activity, tech events, M&A — not static firmographics.
Run lookalikes programmatically with `similar=True` over the API and MCP server.
Why it matters. ICP expansion reflects how companies actually operate today, not how they were classified years ago.
The problem. Filter-based lookalikes match on industry code, headcount band and geography. That says little about how a company operates today.
How it works
Pick 3–5 anchor accounts, such as recent closed-won customers.
The graph analyzes their footprint across capabilities, signals and structural position.
Network-topology matching runs across the full graph.
Ranked results push to your CRM, Clay or sequencing tool over MCP or REST.
Try it (MCP tool call)
```
search_organizations
  similar = True
  anchors = 3–5 entity IDs from your best accounts

# full schema in Docs
```
Related
B2B Relationship Intelligence (Guide · 12 min)
Context Over Quantity: Stop Doing V-Lookups at Scale (Blog · 11 min)
Lookalike
Deeper corporate hierarchy traversal (v1.15)
Released: 2025-10-09
Layer: 1 · Structure
Tags: Identity Graph
For: Sales, RevOps
Where: Product UI · API
Permalink: https://graphiq.ai/resources/changelog/#v1-15 (also #corporate-hierarchy)
Summary. Since v1.15 (October 9, 2025), GraphIQ.ai resolves additional tiers of parent-subsidiary ownership across international entities and shows the full corporate family, including dormant and acquired entities, on account pages.
What shipped
Parent-subsidiary trees now resolve additional tiers of ownership across international entities.
Account pages surface the full corporate family, including dormant and acquired entities.
Why it matters. Reps stop closing a subsidiary blind to the parent — and the broader expansion path it opens.
The problem. A deal with a subsidiary is often a way into its parent and sister companies, but flat account records hide that family.
How it works
Hierarchies resolve from the global ultimate parent down to local subsidiaries and acquired brands.
Acquisitions update the graph, and renamed subsidiaries keep their history.
Account pages now show dormant and acquired entities as part of the family.
Example. Search for Lucasfilm and you get the full 153-entity Disney corporate family, not an isolated row.
Related
If AI Only Had a Brain (Blog · 6 min)
RevOps, A Bloated Tech Stack, and The Knowledge Graph (Blog · 7 min)
Identity Graph
Looking back
2023 — Built graph-first. GraphIQ.ai started with entities and relationships as the foundation, instead of bolting a model onto a flat database.
Oct 2025 — Structure and matching get deeper. Corporate hierarchies resolve further across international entities, and lookalike matching moves from static firmographics to a live graph signature. (v1.15, v1.16)
Nov – Dec 2025 — Fresher facts, agents that act. Employee records stop waiting for quarterly batches, and agents can write what they learn back to the graph. (v1.17, v1.18)
Jan – Feb 2026 — Search by intent. Fingerprint Search reaches general availability, and funding rounds land on the right entity within 24 hours of filing. (v1.19, v1.20)
Oct 2026 — GraphIQ goes agent-native. MCP 2.0 gives any MCP-compatible AI tool 17 read-only tools and six playbooks, and every place in the graph resolves to one canonical name. (Oct 2026)