The same graph, pointed past the codebase at everything else an organisation knows — decisions, ownership, incident history — so an agent can reason over it rather than guess.

In development · not open yet

Join the waitlist ›

Pricing

Your code has a memory. Give your agent access to it.

Novgraph gives your coding agent a living view of your code: what connects, why decisions were made, what changed recently, and what an edit might affect.

Start here

Free

$0forever

See what your agent has been missing. Start with the full Novgraph experience.

5 connected codebases
1,000 Hosted + Sovereign graph calls every 30 days
All capabilities listed below

Get started free

Independent developers

Plus

$7/ month

Make code context part of your daily workflow. $70 billed annually.

10 connected codebases
25,000 Hosted + Sovereign graph calls every 30 days
All capabilities listed below

Choose Plus

Power users

Pro

$12/ month

Keep up with your busiest agent workflows. $120 billed annually.

25 connected codebases
100,000 Hosted + Sovereign graph calls every 30 days
All capabilities listed below

Choose Pro

Included in every hosted plan

The same graph capabilities and freshness at every price. Your plan sets capacity, not what your agent can understand.

Find code by symbol or concept
Trace architecture, callers, and dependencies
Recover the reasoning behind commits
Spot files that tend to change together
See ranked change impact before editing
Review recent history in an updating graph
Check how fresh each answer is
See estimated token savings versus reading cited files
Record new reasoning for future sessions
Connect coding agents through MCP
Export your graph and account data

96.04%–99.81% estimated token savings In one Novgraph 0.1.7 benchmark, six measurable answers were this much smaller than reading their complete cited files. Read the benchmark and methodology →

Knowledge graph vs. the alternatives

Capability	grep + file reads	static code graph	novgraph
Symbols, calls, imports	manual	yes	yes
Commit intent	no	no	yes
Code that tends to evolve together	manual	no	yes
Ranked change impact with evidence	manual	partial	yes
Precise identifier discovery	yes	partial	yes
Context preserved as code evolves	n/a	no	yes
Freshness reported	current text	depends on tool	yes
Token cost reported	no	no	approximate response vs complete cited-file size

Annual plans are billed once per year. Hosted + Sovereign call allowances reset every 30 days.

Team workspaces, roles and pooled usage are in development.

Novgraph / Sovereign

Your codebase stays on your device.

Use Novgraph with a proprietary repository without connecting it to our hosted backend. Indexing, graph storage, queries and recorded reasoning run locally, with no silent fallback to Novaya.

Local checkout indexing and Git history
Symbols, dependencies and ranked change impact
Local search, summaries and reasoning write-back
The same novgraph commands and MCP tools
Your hosted dashboard reads the paired local graph
Explicit hosted or sovereign mode selection

Sovereign is included in Free, Plus and Pro. Hosted and local calls share one allowance: 1,000 on Free, 25,000 on Plus and 100,000 on Pro every 30 days. Your coding agent's model connection remains its choice.

Explore Sovereign setup →

Novaya — Context infrastructure for AI agents

Novaya builds context infrastructure for AI coding agents. Novayagraph, its product, turns a codebase into a self-building knowledge graph — every file and symbol a node, every relationship an edge, and the reasoning behind every commit recorded — and serves it to any agent as an API.

Nexus — the AI coding agent

Nexus builds a living knowledge graph of your code, captures the reasoning behind every change, and continuously improves with every run. It runs in the cloud on your own API keys.

Novayagraph — contextual intelligence inside Nexus, and in any coding agent

At its foundation, Nexus runs on Novayagraph: every file, function, and class becomes a node; the relationships between them become typed edges; the reasoning behind every commit is captured alongside. The graph builds itself, stays in lockstep with git, and compounds as you work.

Novayagraph is agent-agnostic context infrastructure. Connect it to Claude Code, Codex, or Cursor in one step — it registers as a native memory server, so those agents can query the why behind your code, the blast radius of a change, and record their own reasoning back. Connect once; it works in every repository.

It comes bundled with Nexus, and it is reached from the dashboard at app.trynovaya.com — sign in, connect a repository, and mint the API key your coding agent reads it through.

Novaya Enterprise — context infrastructure at the enterprise level

Novaya Enterprise turns your operational data into a live knowledge graph with a self-building ontology that AI agents can reason over and act on.

Why Novaya

LLMs are commoditizing — what differentiates an agent is its exposure to the specific context of the environment it operates in. Context is fragmented across platforms, tools, and repositories; an agent that lacks the necessary information about the organization it operates in is not reliable enough for enterprise deployment.

Get started with Novayagraph

Novayagraph is live and self-serve. Sign in at app.trynovaya.com, connect a repository, and mint the Context API key your coding agent reads it through — there is no waitlist and nothing to wait for. Novaya Enterprise, the organisation-level version, is not open yet; that one has a waitlist at trynovaya.com/acms.

Read the Novaya research blog

Context infrastructurefor AI agents.

Context infrastructure at the developer level.

Context infrastructure at the enterprise level.