Remember what matters.
For the task in front of you.
MCIR stands for Memory Context & Intent Response. Understand the request, bring the relevant history, and give the selected model a useful starting point, not the entire archive.
- Memory
- What should carry forward?
- Context
- What matters for this task?
- Intent
- What are you trying to do?
- Response
- What can we usefully return?
“The release remains a draft until a reviewer signs off.”
A remembered requirement is not evidence that the requirement has been met.01one request, from intent to response
“Resume last week’s release draft.”
An earlier draft, a writing preference, and a review requirement. Follow what joins the request and what stays out.
Start with the task, not the archive.
Resolve what you want to do, which work you mean, and when it happened.
“Resume last week’s release draft.”
- Action
- Resume
- Time
- Last week
- Prior work
- Release brief
Understand. Resume the draft. 3 selected records. Theo Flash.
Fictional records. No model calls or account changes.
Read the complete memory walkthrough
An authored MCIR walkthrough. Fictional records, no live inference, and no account memory changed. Source selection is authored, not a production scoring or contradiction-detection simulation.
- Understand. Resolve what you want to do, which work you mean, and when it happened.
- Recall. Inspect the selected sources and the reason each belongs in this example.
- Assemble. The request and selected context travel together to a qualified model route.
- Respond. A remembered requirement is not proof that the requirement was met.
Resume the draft
Request: Resume last week’s release draft.
Three relevant records in one fictional workspace. The older plan stays out.
Selected sources: draft-014, preference-006, decision-012.
Release brief: draft The retry test is pending and the rollback review remains open. [draft-014] Keep the release in draft until a reviewer signs off. [decision-012] Open question: who will confirm both checks?
Add a new instruction
Request: Resume last week’s release draft. This time, write one paragraph.
The current instruction asks for a paragraph, so this authored example does not apply the older formatting preference.
Selected sources: draft-014, decision-012.
The release is still a draft: the retry test is pending and the rollback review is open. [draft-014] A reviewer’s sign-off is required before approval. [decision-012] Confirm the checks before changing that status.
Try missing context
Request: Resume the release draft from a different project.
None of these records establishes which draft you mean. A source from the wrong project is not a useful substitute.
Selected sources: none.
Which project and release draft should I use? Provide an authorized source or clarify the request before I continue.
Inspect the source records
Last week’s release brief · draft-014 · 2026-09-18. The release brief lists a pending retry test and an open rollback review. This is the earlier artifact the request wants to resume.
How the team writes · preference-006 · 2026-09-17. Use a short summary, then separate confirmed facts from open questions. It shapes the brief’s format, not whether the release is ready.
The review requirement · decision-012 · 2026-09-18. A reviewer must sign off before the release is approved. It preserves a relevant constraint. It does not prove anyone signed off.
An earlier release plan · note-009 · 2026-09-10. An earlier plan proposed releasing before the review meeting. The example leaves this out because a later source requires review. Real conflicts still need validation.
Both illustrated model routes receive the same selected context. The response is authored and does not compare model quality.
Memory is built into supported theo request paths. The full MCIR intent pipeline and Memory Chain have separate activation requirements; confirm availability for your account.
Explore another source-inspection example
Release decision
- Type
- Episodic · decision note
- Origin
- note-014 · message-03
- Recorded
- 2026-09-18
- Scope
- Demo release workspace
“The release remains a draft until a reviewer signs off on the rollback checklist.”
Supports a statement about the review requirement. It does not prove that review has happened.
Prepare the release summary as a draft; reviewer sign-off is still required. [note-014]
Read all three source records
- Release decision · note-014 · message-03
The release remains a draft until a reviewer signs off on the rollback checklist.
Supports a statement about the review requirement. It does not prove that review has happened.
Illustrative draft: Prepare the release summary as a draft; reviewer sign-off is still required. [note-014]
- Writing preference · preference-006 · message-02
Use a short summary, then separate confirmed facts from open questions.
Changes the presentation of a draft. A style preference is not evidence that the release is ready.
Illustrative draft: Summary: release preparation is in progress. Open question: has the rollback checklist been reviewed? [preference-006]
- Earlier release plan · note-009 · message-01
The draft plan proposed releasing before the review meeting.
Conflicts with the later release decision. Inspect the source; do not quietly treat it as current.
Illustrative draft: Do not infer a release date from this superseded plan. Ask the owner to confirm the current decision. [note-009]
Authored example. Synthetic records, no live inference, and no network requests. Choosing a source changes this explanation only; it does not change account memory.
02compare the work you assemble
Not just storage. Context in the orchestration path.
The useful question is not “does it remember?” It is “how does memory reach the work, and what does my team still need to connect?”
Replay chat history
Send earlier turns with the next request. Useful inside a thread, but your app still decides what fits, what remains relevant, and how context moves elsewhere.
You assemble the conversation window.Store and retrieve
Connect a separate service to capture and retrieve context. Your integration determines how those results reach the model and fit with tools and routing.
You connect memory to the rest of the workflow.Orchestrate with theo
Supported paths bring memory together with model routing and skills. MCIR’s intent-driven path prepares a source-linked context handoff; its full pipeline requires activation.
Context and orchestration are designed together.These are integration patterns, not a ranking. Advanced memory services also offer graphs, provenance, and temporal handling. The difference to evaluate is the work your team integrates and operates, not a claim that only one product can remember.
03different context, different roles
Across the work. Within your scope.
A conversation is not the same thing as a memory.
conversation_id links turns within a thread. Built-in recall on supported paths can bring relevant history across threads without a separate memory API. That is distinct from the full, gated MCIR intent pipeline illustrated above.
Supported API and connected-app paths can share owner-scoped context. Memory is scoped to its user, account, or organization, not shared across tenants. Your app must preserve the right authenticated owner and resource permissions; an adapter alone does not enable every memory feature.
The selected model receives relevant context, not the entire archive. Consistent memory does not mean identical answers, automatic access to every application, or retraining a model on your data.
- Episodic
- What happened in a conversation, including a turn or the artifact it produced. Useful when someone refers to “that draft.”
- Semantic
- A distilled fact with a source. Useful context, but still something to validate when the underlying situation changes.
- Procedural
- A preference or working convention: how a person wants a summary structured, for example. Not an authorization to act.
- Memory Chain gated
- Conversation-level synthesis and open-loop context. This layer has separate activation requirements; do not assume it runs for every caller.
04design for inspection
The application still owns the context decision.
- 01
Confirm the owner
Keep conversation IDs and memory access tied to your authenticated user or organization. A reference in a prompt is not a permission grant. - 02
Show the source
Where a result depends on prior context, expose the relevant origin and date. Let the reviewer distinguish a direct source from a model’s summary of one. - 03
Handle contradiction
When a newer decision conflicts with an older note, flag the conflict. Ask for confirmation rather than silently choosing the more convenient version. - 04
Plan retention and deletion
Review which records your workflow creates and how they are retained or removed. Lower recall relevance is not deletion; conversation, audit, and artifact retention are different concerns.
05read the availability labels
Documented, gated, and planned are different states.
- Built-in memory Documented · supported paths
- No separate memory API is required for supported integrations. Ordinary recall is distinct from the full intent-and-envelope pipeline.
- MCIR intent pipeline Implemented / gated
- The intent, source selection, and context handoff shown here have activation and caller-eligibility requirements. Not every completion uses this full path.
- Memory Chain Implemented / gated
- Conversation summaries, open loops, and behavioral context have a separate activation gate. Confirm availability before depending on proactive cross-thread recall.
- Learning and recall upgrades Next / planned
- The published roadmap separates outcome-driven scoring, vector recall, and compression upgrades from the currently available paths.
Inspect before expanding.
Begin with a bounded use case and synthetic records. Check owner scope, stale-source handling, and the difference between remembered context and new evidence before introducing sensitive information.
go deeper into the protocolBuild the next step without starting over.
Bring models, skills, and supported memory paths into one release assistant. Choose a builder-led or engineering-led starting point.