Summary

This PR fixes a memory-plugin architecture flaw: OpenClaw currently has one exclusive plugins.slots.memory owner even though the product, docs, bundled plugins, and runtime behavior already treat “memory” as several different responsibilities.

Today, factual recall, auto-capture, compaction, dreaming/consolidation, and user/session modeling are all forced through the same global selector. That makes documented memory stacks impossible or brittle: selecting one memory provider can unintentionally displace another provider that owns a different responsibility.

This PR corrects that mismatch by introducing purpose-specific memory role slots:

• memory.recall
• memory.capture
• memory.compaction
• memory.dreaming
• memory.userModel

The new role slots are the mechanism. The fix is that OpenClaw can now represent the memory architecture it already documents and partially implements.

What Problem This Solves

This is a flaw fix, not just a feature request.

OpenClaw’s memory ecosystem is already multi-role:

• memory-core / QMD provide local factual recall/search over project memory, session transcripts, and Markdown/QMD corpora.
• Honcho provides cross-session user, agent, and session modeling.
• LanceDB provides embedding-backed long-term recall and auto-capture.
• Active Memory orchestrates recall before generation by using whichever provider owns recall.
• Memory Wiki compiles durable memory into a companion knowledge vault rather than replacing active memory.
• Dreaming/consolidation is a separate runtime responsibility from ordinary recall.

The current plugins.slots.memory contract cannot express that architecture. It treats every kind: "memory" provider as competing for one owner slot, even when the providers are designed to do different jobs. That leads to three concrete problems:

1. Documented combinations are not representable. Configs such as builtin/QMD recall + Honcho user modeling, LanceDB capture + local recall, or Active Memory over a selected recall provider cannot be modeled cleanly through one memory owner.
2. Runtime behavior depends on exceptions and side paths. Plugins that should compose have to rely on context-engine ownership, sidecar behavior, fallback logic, or plugin-specific handling instead of a shared contract.
3. Diagnostics and lifecycle handling become misleading. Startup, validation, status, doctor repair, skill publishing, uninstall cleanup, runtime memory resolution, and SDK-facing aliases need to know which provider owns which memory responsibility.

So while this PR adds new public config keys, the intent is corrective: make the config/runtime contract match the memory system OpenClaw already exposes.

User Impact / Intended Outcome

After this PR, OpenClaw can express memory stacks that were previously documented or implied but not cleanly configurable:

• memory-core or QMD can own factual recall while Honcho owns user/session modeling.
• LanceDB can own recall and/or capture without forcing unrelated memory roles through the same selector.
• Active Memory can use the configured memory.recall provider instead of depending on hard-coded assumptions.
• Dreaming/consolidation can be selected separately from ordinary recall where supported.
• Memory Wiki can remain a companion compiled-knowledge layer beside active memory providers.
• Existing configs using plugins.slots.memory continue to work as recall selection.
• openclaw doctor --fix can migrate/canonicalize non-conflicting legacy memory-slot config while preserving the legacy key for plugin compatibility.

Linked Context

Closes #60572.

Directly related memory-composition requests and prior proposals:

• Feature Request: Support for supplementary/additive memory plugins alongside the primary memory slot #38874: supplementary/additive memory plugins alongside the primary memory slot.
• feat(plugins): allow stackable memory plugins #2306: earlier closed PR proposing stackable memory plugins.
• [Feature]: More memory plugins #2080: earlier request for multiple simultaneous memory plugins and emerging memory architectures.
• Feature Request: Pluggable persistent memory backend for agent context #77714: pluggable persistent memory backend demand.

Dreaming and memory-role separation context:

• [Feature] Decouple dreaming from memory-core — allow any memory plugin to implement dreaming protocol #85473: future protocol work for making dreaming implementable by non-memory-core memory plugins. This PR adds the memory.dreaming slot contract but does not implement that full protocol.
• Allow memory-core dreaming runtime to coexist with third-party memory slot plugins #63067 and fix(plugins): exempt dreaming engine from memory slot fast-path in loader #65411: prior work around keeping dreaming behavior available when another plugin owns the legacy memory slot.
• fix(memory): respect selected slot in dreaming config #62275: previous selected-slot dreaming config behavior.

Runtime/diagnostic slot handling context:

• Memory runtime checks should load the configured memory slot #76567 and fix(memory): load configured memory slot for runtime checks #76495: configured memory-slot runtime checks should load/consult the selected memory plugin.
• [Bug]: openclaw doctor reports 'No active memory plugin is registered for the current config.' when memory-lancedb configured #78540: doctor/status memory diagnostics could misreport configured memory providers.
• Bug: doctor --fix does not clean up plugins.slots.* when plugin is missing, leaving config unrecoverable #76550: stale plugins.slots.* references could leave config unrecoverable until doctor cleanup was broadened.

Context-engine compatibility context:

• [Bug]: lossless-claw selected and enabled but not registered as context engine in 2026.5.2 #77063, [Bug]: lossless-claw context engine not registered in 2026.5.2 (Windows npm global, duplicate plugin id observed) #77185, [Bug]: Bug Report: Context engine plugin not registered despite being loaded #76576, and Bundled legacy context engine not registered in CLI compaction lifecycle worker — dreaming diary subagent fails silently (follow-up to #76576) #81843: context-engine registration and LosslessClaw-adjacent failures shaped the compatibility review for preserving the existing global plugins.slots.contextEngine behavior while rejecting unsupported per-agent contextEngine config.

Future API/RFC context:

• RFC: machine-readable memory-slot ownership — plugins.slots.<slot>.owner #70823: future machine-readable slot ownership/provenance. This PR intentionally keeps scalar plugin IDs and solves the memory role composition layer first.

Memory-plugin interop research: the supporting research document validates this PR as necessary architecture work, not optional extensibility: OpenClaw's own documentation already describes memory as a composable ecosystem where builtin/QMD recall, Honcho user/session modeling, LanceDB recall/capture, Active Memory orchestration, dreaming/consolidation, and Memory Wiki compilation can coexist, but current origin/main collapses every kind: "memory" provider into one exclusive plugins.slots.memory owner, making those documented combinations impossible or dependent on brittle exceptions. PR #88504 closes that product/docs/runtime gap by replacing the single memory owner with explicit role slots — memory.recall, memory.capture, memory.compaction, memory.dreaming, and memory.userModel — while preserving legacy memory as a recall shorthand, so the documented memory-plugin interop story becomes actually representable in config and runtime behavior.

What Changes

• Adds canonical memory role slots: memory.recall, memory.capture, memory.compaction, memory.dreaming, and memory.userModel.
• Keeps legacy plugins.slots.memory as a backward-compatible shorthand for memory.recall during an extended migration window.
• Adds per-agent memory role-slot overrides under agents.list[*].plugins.slots, limited to memory roles plus legacy memory.
• Wires role-slot selection through config schema/types/help/docs, plugin activation, gateway startup preloading, runtime memory resolution, skill publishing, validation, doctor repair/preview, uninstall cleanup, status diagnostics, and SDK-facing aliases.
• Preserves the existing global plugins.slots.contextEngine path for LosslessClaw and other context-engine plugins.
• Explicitly rejects unsupported per-agent agents.list[*].plugins.slots.contextEngine overrides because agent-scoped context-engine semantics are outside this PR.

Implementation Notes

• Adds src/plugins/slot-resolution.ts as the shared role-slot helper layer for canonical memory role keys, defaults, normalization, effective per-agent resolution, and configured selection collection.
• Updates config schema/types/help/labels/docs for global and per-agent memory role slots.
• Keeps PluginSlotsConfig.contextEngine for global plugin slots, and adds an agent-slot type/schema that omits contextEngine.
• Updates doctor legacy memory-slot handling so doctor --fix canonicalizes to memory.recall while retaining legacy memory during the compatibility window.
• Updates plugin loader/activation/startup resolution so explicit role-slot selections activate memory plugins without confusing default memory-core fallback with a user-selected plugin under restrictive allowlists.
• Extends validation to report missing role-slot plugin refs and validate schemas for plugins selected only by non-recall or per-agent memory roles.
• Extends skill publishing, status scanning, uninstall cleanup, memory runtime, dreaming, compaction provider, and SDK alias behavior to use the role-aware model.
• Keeps memory runtime registrations addressable by plugin id, so runtime recall dispatch follows the resolved global or per-agent memory.recall owner.
• Passes the session agent id into Pi embedded extension factory construction so safeguard compaction provider selection honors per-agent memory.compaction overrides in both normal session runs and manual/queued compaction paths.
• Updates bundled memory plugin manifests/docs to describe role support and the recall shorthand migration.

Tests / Verification

Current PR branch:

• Branch: kklouzal:feat/multi-slot-memory-architecture
• Base target: openclaw/openclaw:main
• Latest pushed commit: 9040ece51de5aae8178225cf5d6e8f82dde1886d
• Current PR merge state after push: UNSTABLE.

Latest validation at current pushed head:

• pnpm vitest run test/scripts/plugin-prerelease-test-plan.test.ts: passed before the inheritance repair follow-up.
• node scripts/run-vitest.mjs src/plugins/slots.test.ts src/plugins/uninstall.test.ts src/commands/doctor/shared/stale-plugin-config.test.ts: passed after the inheritance repair follow-up.
• Rebase onto latest origin/main: completed with one conflict in test/scripts/plugin-prerelease-test-plan.test.ts, resolved to the canonical timeoutMs: 1_500_000 expectation.
• git status --short --branch: clean and on pr-88504-ci-remediation...fork/feat/multi-slot-memory-architecture.

Earlier exact-head proof gates carried forward by current head:

• Dev-01 failed CI shard local equivalent: passed, 6 files, 55 tests.
• Dev-01 conflict-focused memory/plugin runtime Vitest: passed, 3 files, 65 tests, 127 skipped.
• Dev-01 pnpm check:test-types: passed.
• Dev-01 pnpm format:check src/plugins/memory-runtime.ts: passed.
• Dev-01 node scripts/check-no-conflict-markers.mjs: passed.
• Dev-01 git diff --check: passed.
• Orion/Nano-01 rollback/upgrade matrix from v2026.5.28 to PR head 3f1f547042220aee42b3e39b8cb7ecb15cfac993: passed for release install/status/valid doctor-fix, expected old-release config rejection of future PR-era slot keys, PR install/config/doctor/status/plugins/models/agents/slots, focused runtime tests, focused doctor/legacy tests, conflict-marker check, and diff check.

Evidence

Rebase/Squash/Proof

• Common upstream base: 13ecca5408cbc94e66105608bfe770d0fc660d76 (origin/main, fix(telegram): back off session init spool retries)
• Previous PR head: 9040ece51de5aae8178225cf5d6e8f82dde1886d
• Current PR head: aaa40f2ad50ca4ab1e5a5171f3de34d844e6fe8f
• Rebased and squashed to one commit on the common base; pushed with --force-with-lease.

ClawSweeper Requested Real Behavior Proof Coverage

ClawSweeper requested redacted real behavior proof for the durable memory role-slot surface before merge: migration/upgrade compatibility, config/schema and plugin loader behavior, memory runtime role-slot selection, doctor migration/repair, status/models output, legacy plugins.slots.memory compatibility, canonical role slots, per-agent overrides, disabled/none recall slots, live CLI agent turns, and gateway service-dispatched live agent execution.

The proof below covers those requested areas with redacted rollback/upgrade output, focused config/doctor/status/plugin-loader gates, live CLI agent-turn markers, exact-head gateway service install/restart/status/health checks, gateway-dispatched live agent execution, and additional automation/plugin-subagent bookkeeping validation. Private host/path/token/session/network values are redacted while preserving commands, markers, artifact names, and commit hashes.

Caveat: the full runtime rollback/upgrade/gateway proof is carried forward from PR head 3f1f547042220aee42b3e39b8cb7ecb15cfac993, while later focused validation covers current pushed head 7fe86ce04ec6df8a571dba1465507ab35be963ba. If the branch changes materially before merge, refresh the exact-head runtime proof as ClawSweeper requested.

Real / Live Behavior Proof

Latest tested PR commit:

head=9040ece51de5aae8178225cf5d6e8f82dde1886d
head_short=9040ece51de5
commit=feat(memory): add multi-slot memory role architecture
branch=kklouzal:feat/multi-slot-memory-architecture
base=openclaw/openclaw:main

Latest validation at this commit:

• PR feat(memory): add multi-slot memory role architecture #88504 was rebased onto the latest origin/main, squashed, then updated with fix commit 7fe86ce6aeddcc55ca41762f16fd19b798f3042d for per-agent memory slot inheritance repair.
• The only merge conflict was scripts/plugin-sdk-surface-report.mjs.
• Conflict cause: origin/main changed SDK surface budgets while this PR branch added memory role-slot SDK surface area.
• Resolution kept both sides and updated aggregate SDK surface budgets to match the merged tree:
  • OPENCLAW_PLUGIN_SDK_MAX_PUBLIC_ENTRYPOINTS: 321
  • OPENCLAW_PLUGIN_SDK_MAX_PUBLIC_EXPORTS: 10338
  • OPENCLAW_PLUGIN_SDK_MAX_PUBLIC_FUNCTION_EXPORTS: 5187
  • OPENCLAW_PLUGIN_SDK_MAX_PUBLIC_DEPRECATED_EXPORTS: 3246
• Focused validation at 7fe86ce6aeddcc55ca41762f16fd19b798f3042d: node scripts/run-vitest.mjs src/plugins/slots.test.ts src/plugins/uninstall.test.ts src/commands/doctor/shared/stale-plugin-config.test.ts passed.
• Merge conflict verification after remediation: git merge-tree --write-tree HEAD origin/main completed cleanly with no conflict output.
• Worktree after remediation: clean.

Earlier exact-head proof carried forward by the merge commit:

• CI remediation commit c447e22033ca104cb9d5292edf4313bcb74f81cf fixed the PR-owned check-prod-types failure by removing the now-unused active-memory config parameter/helper after the memory tool allowlist behavior change.
• Verified on Dev-01 with pnpm tsgo:prod passing before push.
• Full proof refresh at 3f1f547042220aee42b3e39b8cb7ecb15cfac993 covered memory role-slot contract, per-agent overrides, legacy plugins.slots.memory compatibility, disabled none recall slots, doctor migration/repair behavior, role-slot-selected plugin activation/validation, upgrade behavior from the last public release, exact-head gateway service install/restart, gateway health, gateway-dispatched live agent execution, and plugin subagent completion bookkeeping.

Proof environment:

• Dev-01 canonical checkout.
• Orion/Nano-01 Linux rollback/upgrade test box.
• Nano was tested at PR head 3f1f547042220aee42b3e39b8cb7ecb15cfac993, rolled back to public release v2026.5.28, then upgraded directly back to the same PR head.
• Runtime logs were redacted for private host, path, token, session, and network values.

validated_current_head=9040ece51de5aae8178225cf5d6e8f82dde1886d
full_runtime_proof_head=3f1f547042220aee42b3e39b8cb7ecb15cfac993
release_base=v2026.5.28 e93216080aa1f425d3ab127014603eba8e365b2d
nano_proof_bundle=proof-full-2b0d657729a3-20260604T025709Z
nano_auth_backup_sha256=eef060cae6f9f53141bdcbb5fb45960cb18c935a4d5e094599a9bac4fc62a1c3
dev01_gateway_artifact=.artifacts/proof-full-2b0d657-dev01-20260604T030012Z
upgrade_path=auth backup -> exact-head baseline -> release checkout/install/doctor/status/models -> PR checkout/install/config/doctor/status/models/slots/agents -> post-upgrade live provider turn -> exact-head gateway service force install/restart -> gateway live agent dispatch

Dev-01 exact-head deterministic gateway/plugin-subagent gates:

- src/gateway/server-plugins.subagent-ended-hook.test.ts
- src/gateway/server-plugins.test.ts
- src/gateway/server-methods/agent.test.ts
- result: passed, 4 files, 337 tests
- coverage focus: plugin runtime gateway dispatch scope, generated idempotency fallback, plugin_subagent tracking, plugin owner propagation, registry registration, completion bookkeeping, retry dedupe, and fallback behavior when registry persistence fails

Nano auth-preserving rollback/upgrade matrix:

- private auth backup before rollback: created, sha256 eef060cae6f9f53141bdcbb5fb45960cb18c935a4d5e094599a9bac4fc62a1c3
- checkout v2026.5.28: passed
- release pnpm install --frozen-lockfile: passed
- release version: passed
- release config validate against PR-era role-slot config: failed as expected because old release does not know future memory role-slot keys
- release doctor --fix --non-interactive: passed
- release status: passed
- release models status: passed
- checkout PR head 3f1f547042220aee42b3e39b8cb7ecb15cfac993: passed
- PR pnpm install --frozen-lockfile: passed
- PR version: passed
- PR config validate: passed
- PR doctor --fix --non-interactive: passed
- PR status: passed
- PR models status: passed; provider auth status ok, missingProvidersInUse=[]
- PR plugins.slots inspection: passed
- PR agents.list inspection: passed
- post-upgrade live main agent turn: passed

Nano clean live source CLI turns after re-auth:

- main: exit=0, marker=OKM2B0D
- honcho-agent: exit=0, marker=OKH2B0D
- dreamer-agent: exit=0, marker=OKD2B0D
- split-agent: exit=0, marker=OKS2B0D
- disabled-recall-agent: exit=0, marker=OKN2B0D

Nano exact-head gateway service proof:

- gateway install --force: passed
- gateway restart: passed
- gateway status: passed; OpenClaw CLI version 2026.6.9, Gateway version 2026.6.9, runtime running, connectivity probe ok, capability write-capable, listening on loopback port 18789
- gateway health: passed, OK
- gateway stability: passed
- gateway call agent --expect-final through running service: passed, status=ok, marker=LIVE_GATEWAY_AGENT_FORCE_2B0D

Nano focused exact-head source gates after upgrade:

- focused memory/plugin runtime Vitest: passed, 4 files, 245 tests
- focused doctor/legacy slot Vitest: passed, 3 files, 81 tests
- check-no-conflict-markers: passed
- git diff --check: passed
- final git status: clean for tracked files; only pre-existing generated `1/` build-output directories remained untracked from doctor/UI build paths

Verified role-slot matrix:

pr_global_slot_matrix_verified:
- plugins.slots.memory.recall = memory-core
- plugins.slots.memory.dreaming = memory-core
- plugins.slots.memory.userModel = openclaw-honcho
- plugins.slots.contextEngine = lossless-claw

pr_agent_slot_matrix_verified:
- main: no explicit per-agent memory override
- honcho-agent: memory.recall = openclaw-honcho; memory.userModel = openclaw-honcho
- dreamer-agent: memory.recall = memory-core; memory.dreaming = memory-core; memory.userModel = openclaw-honcho
- split-agent: memory.recall = memory-core; memory.capture = active-memory; memory.userModel = openclaw-honcho
- disabled-recall-agent: memory.recall = none; memory.capture = active-memory; memory.userModel = openclaw-honcho

Observed result:

• Current pushed PR head is 9040ece51de5aae8178225cf5d6e8f82dde1886d.
• The current head includes the full memory role-slot implementation, the latest origin/main rebase, and the squashed PR history.
• The current head passed focused plugin prerelease plan validation after the rebase/squash.
• The earlier full runtime proof demonstrates that the role-slot implementation validates and doctors the intended memory role-slot matrix after direct upgrade from the last public release.
• Auth survived rollback/upgrade testing and was confirmed by post-upgrade provider status plus live agent turns.
• Five clean source CLI agent turns and one exact-head gateway service-dispatched agent turn completed successfully.
• Gateway service proof required gateway install --force; a non-forced install left the already-installed unit in place, which was explicitly caught and corrected before claiming gateway proof.

What was not tested:

• Maintainer taxonomy approval is still a review decision, not something this proof can self-certify.
• Runtime proof logs were redacted to avoid private host, path, token, session, and network values.

Proof limitations or environment constraints:

• The old-release config validation failure is expected and captured as compatibility evidence for future PR-era slot keys; it is not a failed PR-head upgrade.
• The gateway service was loopback-bound on Nano and tested through local gateway CLI calls.

Compatibility and Risk

Backward compatibility preserved:

• Existing plugins.slots.memory configs continue to resolve as recall selection.
• doctor --fix keeps the legacy plugins.slots.memory key in place after copying to memory.recall, so third-party plugins that still read only the legacy key do not hit an immediate flag day.
• The default recall plugin remains memory-core unless explicitly disabled or overridden.
• plugins.enabled = false, denylist behavior, and explicit plugins.entries.<id>.enabled = false remain respected.
• Global plugins.slots.contextEngine remains separate from memory roles and continues to select context engines such as LosslessClaw.

Compatibility changes reviewers should inspect:

• memory.recall, memory.capture, memory.compaction, memory.dreaming, and memory.userModel become durable public config keys after merge.
• Per-agent plugin-slot overrides are memory-only. agents.list[*].plugins.slots.contextEngine is intentionally rejected until a future PR defines and implements agent-scoped context-engine semantics.
• Legacy plugins.slots.memory is still deprecated as a canonical authoring target even though it remains supported and retained for compatibility.

Primary risks:

• The public memory role taxonomy becomes API surface after merge, so naming and docs need maintainer signoff.
• Role-gating mistakes could attach the wrong memory provider or suppress expected memory behavior; this PR adds focused coverage across loader, validation, startup, runtime, doctor, status, skill publishing, and cleanup surfaces.
• Third-party plugins should eventually read memory.recall, but the retained legacy key avoids forcing that ecosystem update immediately.

PR/Commit Surface Breakdown

Source: 67 files, +2216 / -782
Tests: 37 files, +2639 / -98
Total: 104 files, +4855 / -880

Reviewer Checklist

• Review role-slot resolution/default behavior and backward compatibility for legacy plugins.slots.memory configs.
• Review documentation/schema/UI metadata updates against the new canonical memory role slots.
• Confirm the preserved verification and rebase proof cover the current branch head.

Maintenance update (2026-06-25): rebased/squashed cleanup to head 60783bc0a577dc5201210da488f28eed70475df6; removed duplicated gateway watch CI/test remediation hunks from standalone PR #95473 so that #95473 remains the sole carrier of that fix.

Clawsweeper remediation update — 2026-06-26

• Old head: 60783bc0a577dc5201210da488f28eed70475df6
• New head: 4e4d3b190ea2c510ce51e0dd5dbd27296c6e95a3
• Remediated valid source finding: native harness scoped plugin preload now resolves the canonical/per-agent memory.recall role slot instead of only the legacy scalar memory slot.
• Proof: added focused regressions for canonical plugins.slots["memory.recall"] and per-agent memory.recall overrides in src/agents/harness/runtime-plugin.test.ts.
• Validation: git diff --check HEAD^..HEAD, pnpm exec oxlint src/agents/harness/runtime-plugin.ts src/agents/harness/runtime-plugin.test.ts, and node scripts/run-vitest.mjs run --config test/vitest/vitest.agents-support.config.ts src/agents/harness/runtime-plugin.test.ts --reporter verbose all passed.