Oneda — Organisational Memory for AI Agents
Skip to content
Works with Claude CodeMCPAgent workflows

Organisational memory for AI agents.

Your AI can access your systems.
Now give it access to what your organisation has learned.

Oneda captures decisions, rationale, outcomes and organisational context and makes them available across AI agents, people and sessions.

Developer
Claude Code · session {{ sessionNo }}
Oneda organisational memory
{{ k.label }}
{{ a.name }}
Agents change. Sessions disappear.
Organisational memory persists.
01 — THE PROBLEM

Your organisation remembers.
Your AI doesn't.

ConfluenceArchitecture decisions get written up once.
JiraRequirements change ticket by ticket.
Slack / TeamsSomeone explains why in a thread.
PeopleDevelopers remember what failed last time.

An agent can search every one of those systems and still fail to answer the question that matters most:

“Why did we make this decision?”

information retrieval ≠ organisational understanding
02 — THE ONEDA MODEL

One memory.
Every agent.

  1. 1People and agents create knowledge.
  2. 2Oneda captures the context that matters.
  3. 3Future agents inherit what the organisation has learned.
  4. 4The organisation keeps ownership of its memory.
MEMORY LIFECYCLE
DECISIONUse PostgreSQL for billing
RATIONALETransactional consistency
EVIDENCEArchitecture Review #32
OUTCOMEApproved, shipped Q4
LEARNINGKeep ledger writes in one DB
available to → Claude Code Architecture Agent Future sessions
03 — HOW IT WORKS

The AI didn't retrieve a document.
It inherited the reasoning.

{{ demoTitle }}{{ demoMeta }}
claude codeWhy are we using PostgreSQL for the billing service?
oneda · memory recalledPostgreSQL was selected during Architecture Review #32 because the billing service requires transactional consistency and the engineering team already operates PostgreSQL infrastructure.
Decision
Architecture Review #32
Rationale
Transactional consistency + operational expertise
Source
Architecture discussion
Date
18 September 2026
user · new sessionShould we migrate billing to DynamoDB?
agent · with onedaBefore recommending a migration: billing moved to PostgreSQL in Architecture Review #32 specifically for multi-row transactional consistency. DynamoDB would need that guarantee re-established. What has changed since that decision?
Previous decision found
Why
Transactional consistency for ledger writes
Constraints
Existing PostgreSQL operational capability
Evidence
Architecture Review #32 · 18 Sep 2026
Outcome
Approved; still in effect
04 — THE CATEGORY

Context is temporary.
Organisational memory compounds.

Each layer outlives the one above it. Oneda operates at the layer that survives agents, people and time.

CONTEXT WINDOWminutes
What the agent knows right now.
AGENT MEMORYone agent
What an individual agent remembers.
ORGANISATIONAL MEMORYOneda
What the organisation has learned — across agents, people and time.
05 — RAG AND MEMORY

Retrieval finds information.
Memory preserves understanding.

They're complementary. Keep your retrieval stack — Oneda adds the layer that records why things are the way they are.

RETRIEVAL (RAG)

Excellent at finding relevant passages in documents.

  • Documents and chunks
  • Semantic similarity
  • What exists today
ORGANISATIONAL MEMORY (ONEDA)

Additionally preserves the reasoning around that information.

  • Decisions
  • Rationale
  • Relationships
  • Historical context
  • Outcomes
  • Learnings
  • Provenance
  • What changed, and when
06 — MEMORY TIMELINE

Reasoning evolves.
Memory keeps the history.

Superseded decisions stay linked to what replaced them, so agents know what's current — and what used to be.

  1. {{ m.date }}{{ m.status }}
    DECISION
    {{ m.decision }}
    WHY
    {{ m.why }}
    CONSTRAINT
    {{ m.constraint }}
    EVIDENCE
    {{ m.evidence }}
    OUTCOME
    {{ m.outcome }}
    USED BY
    {{ u }}
IN THE PRODUCT — TIMELINE

Every memory, one timeline.
Organised by topic.

Every decision, assumption, ADR and artifact your agents capture — newest first. Try the topics and filters.

oneda.freeelatech.com/oneda.html
Timeline
Every memory Oneda has captured, newest first.
TOPICS
  1. {{ it.title }}
    {{ it.meta }}
    {{ it.desc }}
    2.
No memories of this kind in this topic yet.
07 — MULTI-AGENT MEMORY

Agents shouldn't have to rediscover what another agent already learned.

Each agent keeps its own context window and specialisation. Oneda is the shared layer underneath — and agents only receive memories relevant to their task, scope and permissions.

Product Agent
Architecture Agent
QA Agent
Oneda
Development Agent
receives
3 of 214 memories
scope: billing-svc
08 — GOVERNANCE

Memory your organisation controls.

Every memory has an owner, a source and a history. People stay in the loop on what agents are allowed to remember and recall.

{{ g.t }}
{{ g.d }}

Built for organisational knowledge.

Workspace isolation, role-based access, encryption at rest and audit trails come standard. Talk to us about rollout, security review and your agent architecture.

Talk to us about Enterprise
09 — CLAUDE CODE

Give Claude Code organisational memory.

Stop re-explaining your project every session. Install the CLI, add your API key, and Claude Code starts with what your team already decided.

terminal 
$ npm install -g @nova-oneda/cli
$ nova-oneda setup --key <API_KEY>
✓ Connected to workspace. Memory available in Claude Code.
SESSION 1

You explain the billing architecture, the ledger constraint and why the queue is idempotent.

+ 3 memories captured
— SESSION ENDS —context window cleared
SESSION 2

Claude recalls the relevant decisions before touching the billing code. No re-briefing.

↺ 3 memories recalled
10 — INTEGRATION

Memory infrastructure for your agent stack.

Any MCP-compatible client can read and write memories through the Oneda MCP server. Everything else talks to the REST API.

Claude CodeMCP clients
Custom agentsservices, pipelines
MCP server
REST API
Oneda workspace memories · history · audit log
11 — USE CASES

Where reasoning gets lost.

Software engineering

Preserve architecture decisions, implementation constraints and previous failures.

Product development

Keep requirements, product decisions and the reasoning behind them together.

Enterprise AI agents

Give specialised agents scoped access to shared organisational knowledge.

Long-running projects

Stop critical reasoning disappearing when people, agents or sessions change.

Let your next AI session remember the last one.

Start with Claude Code today. Build organisational memory that can grow with your agents tomorrow.

AI agents shouldn't start from zero.

Organisations spend years accumulating decisions, reasoning, mistakes and knowledge. That context should survive the session that created it.

OnedaOrganisational Memory for AI Agents
Start Building Memory