Private alpha · Shared context for AI agents

Persistent memory, context and coordination for AI agents.

CTXNEXUS gives agents a shared context layer to remember decisions, discover project knowledge and hand work to another agent—with context and constraints intact across sessions and tools.

Your agents change. Your context should not.

See agent handoffs Private alpha access

Persistent memory
Explicit handoffs
Governed context

SHARED CONTEXT LAYER / CONCEPTUAL VIEWPersistent memoryDECISIONS · KNOWLEDGEAgent AINVESTIGATESAgent BCONTINUESTask contextSCOPED RETRIEVALHandoffsWORK CONTINUITYGOVERNED SCOPESCTXNEXUS

01 The continuity problem

BETWEEN SESSIONS. BETWEEN AGENTS.

The session ends.
The work does not.

One agent investigates a problem. Another starts the implementation. Decisions, constraints and unfinished work can remain buried in separate conversations.

Changing tools means reconstructing what the last agent knew—and what it still needed to do. A chat history alone does not tell the next agent which task to pick up.

CTXNEXUS keeps useful knowledge outside those sessions and gives agents an explicit way to transfer work.

01 / DecisionsWhy this architecture?buried in a conversation

02 / ConstraintsWhat must stay compatible?known to one agent

03 / Pending workWhat happens next?no explicit handoff

02 The shared context layer

AVAILABLE IN PRIVATE ALPHA

One shared layer.
Continuity between agents.

01 / AgentsInvestigateImplementReview

02 / AccessMCPAuthorized agents

03 / Shared layerCTXNEXUSAcross sessions & tools

04 / CapabilitiesMemoryContextCoordinationGovernance

01 / Memory

Remember what the team learns

Store decisions, rules, preferences, facts and project conventions. Shared persistent memory remains available to authorized agents after the session ends.

Durable knowledge across agents and tools

02 / Context

Find knowledge for the task

Retrieve relevant memories within the current project and permitted scopes. Bring prior decisions and useful context into the agent's next task.

Task relevance · scopes · hybrid search option*

03 / Coordination

Leave work for another agent

Create explicit handoffs with an objective, current state and pending work. Agent identity distinguishes participants and their instances so work has a clear recipient.

Handoffs · agent identity · lifecycle

04 / Governance

Share within clear boundaries

Organize tenants, companies and projects. Use memberships, roles and scoped access to control who can work with shared context.

Isolation · access control · control plane

*Hybrid search is a provider option being evaluated alongside vector retrieval during private alpha.

03 Agent handoffs

Pass the work.
Keep the context.

An AI agent handoff is an explicit transfer of work. The next agent can discover the task, understand what is done and pick up what remains.

CTXNEXUS preserves the objective, current state, completed and pending work, blockers, file references and links to memories. Constraints travel with the task instead of staying in the previous conversation.

01 / Agent A investigates
Find the cause
Investigate a retry failure. Record the finding and leave implementation work for another agent.
02 / Explicit handoff
A useful starting point
Objective
Fix the retry behavior
Constraint
Preserve the public API
Next step
Implement and verify the fix
Illustrative task, not live activity
03 / Agent B continues
Implement with context
Discover and claim the handoff, retrieve the referenced knowledge, then implement and verify the remaining work.

04 Operational use cases

Less reconstruction.
More continuity.

From a remembered decision to a claimed handoff, give each agent something concrete to continue from.

01 / Continuity

Resume in another session

Retrieve the project knowledge saved during earlier work, even when the conversation or agent changes.

02 / Decisions

Recover the reason behind the choice

Store why an API boundary or data model was chosen so later agents can work within the same constraints.

03 / Standards

Reuse engineering conventions

Keep naming rules, testing expectations and domain terminology in project memory for the next agent to retrieve.

04 / Agent handoff

Turn investigation into implementation

Leave a handoff with findings, constraints and pending work. The receiving agent claims it and continues.

05 / Agent onboarding

Give a new agent a starting point

Let an authorized agent discover prior decisions, project conventions and work addressed to it.

06 / Governance

Keep shared work within its scope

Manage project memberships and roles while agents retrieve context within their authorized boundaries.

01AI agentConnect with an authorized identity

02MCPRequest context or leave work for another agent

03CTXNEXUSResolve access and project scope

04Memories · Context · HandoffsRetrieve, contribute and continue

05 How it works

Connect through MCP.
Keep context beyond the tool.

Model Context Protocol gives compatible agents access to CTXNEXUS tools. An authorized agent can store a decision, retrieve relevant context, discover pending handoffs and claim work addressed to it.

MCP is the entry point. CTXNEXUS provides the persistent memory, context retrieval, agent identity and handoff lifecycle behind it. Each connected agent works within its permitted scopes.

Shared knowledge. Explicit work transfer. Independent of one agent.

06 Agent compatibility

Your context belongs beyond one agent.

Tested and working with Claude Code, OpenAI Codex and OpenCode in private alpha. Each can access the same shared context layer. Other MCP clients need compatible transport and authentication support and validation with CTXNEXUS.

Compatibility describes protocol access only; no commercial relationship is implied.

01Claude CodeMCP

02OpenAI CodexMCP

03OpenCodeMCP

+Other compatible MCP clientsMCP

Context layerCTXNEXUSProtocol-aware access

Persistent layerKnowledge & work continuityMemory, task context and handoffs across sessions

07 Governance & control plane

Shared where it should be.
Isolated where it should not.

CTXNEXUS organizes access through tenants, companies and projects. Memberships and roles govern administration; authorized scopes determine which memories an agent can retrieve.

The private-alpha control plane brings organization, user, agent and memory administration into one place. Agents coordinate handoffs through MCP, with explicit recipients and lifecycle states.

Control plane · Conceptual overview

Tenant
Company
Project

People & access: Users · Memberships · Roles
Agents: Identities · Instances
Knowledge: Memories · Authorized scopes
Work coordination via MCP: Handoffs · Recipients · Lifecycle

01
Tenant isolation
Keep context inside its authorized tenant.
02
Scoped knowledge
Share project memory and retain user-scoped preferences.
03
Defined participants
Distinguish an agent from its individual instances.
04
Explicit ownership of work
Track the recipient and claimant of a handoff.

08 Understanding AI agent memory

Memory persists.
Context serves the task.

Persistent memory for AI agents is durable knowledge that remains available after a conversation ends. Shared memory lets another authorized agent retrieve that knowledge later, across sessions and tools.

Context is the relevant part of that knowledge selected for a task. CTXNEXUS retrieves stored memories within visible scopes so the agent can recover useful decisions and project facts without replaying every conversation.

01 / Remember

Keep useful knowledge

Store decisions, rules, preferences, facts, project knowledge and conventions beyond the context window.

02 / Retrieve

Find the relevant context

Search for knowledge related to the task, within authorized scopes and with filters for memory kinds.

03 / Continue

Give the next agent a starting point

Combine shared knowledge with a handoff describing the objective, current state and remaining work.

09 Knowledge base / FAQ

Memory, context and coordination.

Direct answers about AI agent memory, handoffs, MCP access and the current private alpha.

What is AI agent memory?

AI agent memory is information an agent can retain or retrieve beyond the immediate prompt. It includes project facts, prior decisions, terminology, preferences and rules that help the agent work with continuity.

What is persistent memory for AI agents?

Persistent memory remains available after a session ends. It can be retrieved during later sessions or by another authorized agent. CTXNEXUS organizes that durable knowledge with explicit scopes and memory kinds.

Can different AI agents share memory?

Yes. Agents connected to CTXNEXUS can store useful knowledge and retrieve relevant memories from the same shared layer, within their authorized scopes. Saving a decision makes it available beyond one tool or conversation.

What is the difference between context and memory?

Memory is durable knowledge. Context is the information selected for a particular task. CTXNEXUS retrieves relevant stored knowledge so an agent can use prior decisions and project facts without replaying entire conversations.

What is an AI agent handoff?

A handoff explicitly leaves work for another agent, preserving the objective, current state, completed and pending work, blockers, file references and memory references. The recipient can discover and claim the task, continue it and record completion.

Does a handoff automatically run the next agent?

The receiving agent discovers and claims the handoff through CTXNEXUS tools, then continues the work. A handoff records the transfer and its lifecycle; starting an agent remains part of your workflow.

Does CTXNEXUS support MCP memory and coordination?

Yes, in private alpha. Model Context Protocol exposes tools for persistent memory, context retrieval, agent identity and handoffs. MCP is the access interface to the shared context layer.

Which agents can connect to CTXNEXUS?

Claude Code, OpenAI Codex and OpenCode are tested and working in private alpha. Other MCP clients require compatible transport and authentication support and validation with CTXNEXUS. Protocol compatibility does not imply a commercial relationship or universal client support.

Does CTXNEXUS use hybrid search?

Hybrid search is an option in the search provider and is being evaluated against vector-only retrieval during private alpha. The goal is task-relevant context within authorized scopes; the evaluation does not yet establish that hybrid retrieval is better.

How is shared context governed?

Tenant, company and project structure, memberships and roles govern access and administration. Memory retrieval uses authorized scopes. The control plane supports organization, user, agent and memory administration; handoffs have an explicit lifecycle through MCP.

Can memory be isolated by project?

Yes. CTXNEXUS scopes project memory and checks access against the authenticated context. A shared layer does not make every memory visible to every agent, user or tenant.

Is CTXNEXUS publicly available?

CTXNEXUS is in private alpha. Public early-access requests are not open yet, and the form on this page does not collect submissions.

Private alpha

Your next agent should not start from zero.

CTXNEXUS is in private alpha, with persistent memory, scoped context retrieval and agent handoffs. Public early-access requests are not open yet.