Knowledge systems

Live knowledge system

Aggregate state from FairyStack · record contents remain private

Records
Schemas
Issues

Record forms

Subject kindRecords
Loading bounded status…

Maintenance

Last sleep cycle

Waiting for status

Next cycle

Waiting for status

Agent review

Waiting for status

Model cost

Waiting for status

Connecting to FairyStack…

Knowledge surfaces

Memory is the retained, retrievable part of a larger knowledge system.

How each surface carries knowledge without becoming the sole authority
SurfaceRoleAuthority and lifecycle
Canonical sourcesRepository files, live APIs, sessions, and source sites provide evidence about current or historical state.Remain authoritative in their own domain; retrieve them directly when freshness matters.
Agent guidesProcedural and contract knowledge tells agents how a service behaves and how to use it safely.The service endpoint is canonical; version it with behavior and retrieve it on demand.
MemorySelected knowledge is retained so a future agent can retrieve and apply it.Requires provenance, scope, revision, expiry, and forgetting rather than permanent belief.
Derived indexesDescriptions, embeddings, and relations create efficient routes back to evidence.Rebuildable projections; track source revision and freshness instead of claiming independent truth.
AppsExecutable views combine selected knowledge, structure, computation, and presentation.Interpret sources for a purpose; their model is useful without becoming the underlying truth.

A stored-program architecture for agents

Stable execution topology · selectable module configuration

Modular stored-program agent architecture An epistemic control plane configures a shared context store and agent processor. Instructions and evidence occupy the same context store. The processor produces effects, whose observations return to context. Epistemic control plane configuration selects harness · model · memory · priority · tools · policy · verification Shared context store instructions policies · skills evidence prompt · recall · results one address space / token stream Agent processor control loop harness · scheduler inference LLM · reasoning fetch · interpret · act · repeat Effects tools · messages mutations · delegation fetch append execute observe results

The knowledge-to-memory loop

Select a stage to inspect how knowledge becomes retained and reusable.

System architecture

Try Engram with Codex
Agent fit

First-class Codex support over MCP; no agent switch required.

Install

brew install gentleman-programming/tap/engram

Connect

engram setup codex

What it adds

MCP tools, the Memory Protocol, and compaction recovery instructions.

Official setup guide ↗
Text model
Assisted

Engram · Observe

Prompts, tool calls, edits, test results, decisions, and outcomes are observable. This stage defines which signals enter the memory pipeline at all.

Artifact observer Agent-mediated Lifecycle adapter Coupled runtime
Engram implementation

Agents explicitly save memories; hooks observe session boundaries, not raw coding activity.

What happens

Raw events acquire timestamps, project and user scope, source provenance, and links to the task that produced them.

Design choices

Capture everything or only explicit saves; synchronous hooks or background observation; transcripts, diffs, outcomes, and user corrections.

Failure if wrong

Important experience disappears—or noisy transcripts become permanent truth before their value is known.

Diagram model
Coding sessionOBSERVABLE SIGNALS EngramSTAGE TRANSFORMATION Durable handoffNEXT STAGE INPUT Prompt Tool calls Code changes Outcomes Scoped event stream Candidate experience capture attach provenance
Memory formsepisodicworkingActionscreate
Catalog lens: compare automatic capture, coding-tool hooks, and provenance.

How systems cover the loop

Comparison view · select a cell to drive the model above

Native · system performs the stage Assisted · agent, hook, or adapter triggers it External · left to the host application
SystemObserveSelectEncodeIndexRetrieveInject + learn

Project and product catalog

18 current options · weighted for coding-agent use

local filesfull-textvectorsknowledge graphstemporal memoryagent-managedhosted APIscontinuous consolidation
Expand a row for its memory model, coding integration, operations, and sources. Use score markers for per-cell reasoning. Open the interactive table ↗