feat(daemon): inherit parent context for sub-agent sessions

Author: NicholaiVogel

docs/API.md

@@ -2386,12 +2386,22 @@ for injection into the harness system prompt. Requires `remember` permission
 ```json
 {
   "harness": "claude-code",
+  "project": "/workspace/repo",
+  "agentId": "optional-signet-agent-id",
+  "harnessAgentId": "optional-harness-subagent-id",
+  "parentSessionKey": "optional-parent-session-key",
   "sessionKey": "session-uuid",
   "runtimePath": "plugin"
 }
 ```
 
-`harness` is required.
+`harness` is required. `agentId` is the Signet persistence scope. Harness
+native sub-agent identifiers, such as Claude Code's `agent_id`, must be sent as
+`harnessAgentId`; they are lineage hints and are not used for Signet data
+scoping. `parentSessionKey` may be provided when the harness exposes explicit
+lineage. If it is absent, Signet infers parent context where possible from
+harness-native signals such as OpenClaw lineage session keys or recent Claude
+Code parent activity in the same project.
 
 **Response** — implementation-defined context object returned by
 `handleSessionStart`.
@@ -2774,6 +2784,69 @@ Both raw keys (`abc123`) and prefixed keys (`session:abc123`) are accepted.
 
 Returns `404` if the session key is not found.
 
+### GET /api/sessions/:key/transcript
+
+Return the canonical cleaned transcript for a session. Results are scoped to
+the authenticated agent; pass `agent_id` only when calling with an authorized
+agent scope.
+
+Both raw keys (`abc123`) and prefixed keys (`session:abc123`) are accepted.
+
+**Response**
+
+```json
+{
+  "sessionKey": "session-uuid",
+  "agentId": "default",
+  "content": "User: ...\nAssistant: ..."
+}
+```
+
+Returns `404` if no transcript exists for that session and agent scope.
+
+### POST /api/sessions/search
+
+Search active or completed session transcripts. This route powers the
+`session_search` MCP tool and is intended for sub-agents that need to inspect
+the parent session without forcing a large token snapshot into every spawn.
+Results are agent-scoped and require `recall` permission.
+
+**Request body**
+
+```json
+{
+  "query": "Juniper trunk ports",
+  "sessionKey": "optional-specific-session",
+  "currentSessionKey": "agent:nicholai:subagent:abc123",
+  "agentId": "nicholai",
+  "project": "/workspace/repo",
+  "limit": 5
+}
+```
+
+`query` is required. `limit` is clamped to `1..20`. If `sessionKey` is absent
+and `currentSessionKey` encodes OpenClaw sub-agent lineage, Signet defaults the
+search to the inferred parent session. Otherwise, Signet searches transcripts
+in the requested agent and project scope while excluding `currentSessionKey`.
+
+**Response**
+
+```json
+{
+  "query": "Juniper trunk ports",
+  "hits": [
+    {
+      "sessionKey": "agent:nicholai:main",
+      "project": "/workspace/repo",
+      "updatedAt": "2026-03-25T10:05:00.000Z",
+      "excerpt": "keep the Juniper EX4300 VLAN audit focused on trunk ports",
+      "rank": -1.2
+    }
+  ],
+  "count": 1
+}
+```
+
 ### GET /api/sessions/summaries
 
 List temporal summary nodes used for drill-down and `MEMORY.md` synthesis.
@@ -3852,9 +3925,9 @@ MCP Server
 Model Context Protocol endpoint using Streamable HTTP transport (stateless).
 Supports POST (send messages), GET (SSE stream), and DELETE (session teardown).
 
-Exposes memory tools: `memory_search`, `memory_store`, `memory_get`,
-`memory_list`, `memory_modify`, `memory_forget`. See `docs/MCP.md` for full
-tool documentation.
+Exposes memory/session tools: `memory_search`, `session_search`,
+`memory_store`, `memory_get`, `memory_list`, `memory_modify`, `memory_forget`.
+See `docs/MCP.md` for full tool documentation.
 
 **POST /mcp** — Send MCP JSON-RPC messages. Returns JSON or SSE stream.
 

docs/CONFIGURATION.md

@@ -916,6 +916,29 @@ Checkpoints are triggered by five events: `periodic`, `pre_compaction`,
 storage.
 
 
+### Sub-agents (`subagents`)
+
+Controls deterministic parent-session context inherited by sub-agent sessions
+at `session-start`. This uses stored active transcripts and checkpoints; it
+does not make an LLM call.
+
+| Field | Default | Range | Description |
+|-------|---------|-------|-------------|
+| `inheritContext` | `true` | — | Inject a compact parent context block when parent lineage is available |
+| `tailChars` | `3000` | 0-20000 | Max transcript tail characters included from the parent session |
+
+```yaml
+memory:
+  pipelineV2:
+    subagents:
+      inheritContext: true
+      tailChars: 3000
+```
+
+Set `inheritContext: false` to disable automatic inherited context while
+leaving the explicit `session_search` MCP/API surface available.
+
+
 ### Telemetry (`telemetry`)
 
 Anonymous usage telemetry. Only active when `telemetryEnabled: true`.

docs/HARNESSES.md

@@ -227,8 +227,9 @@ Claude Code also gets native MCP tool access to Signet memory via the
 }
 ```
 
-This gives Claude Code direct access to `memory_search`, `memory_store`,
-`memory_get`, `memory_list`, `memory_modify`, and `memory_forget` tools.
+This gives Claude Code direct access to `memory_search`, `session_search`,
+`memory_store`, `memory_get`, `memory_list`, `memory_modify`, and
+`memory_forget` tools.
 
 ### Prerequisites
 
@@ -269,7 +270,7 @@ lineage.
 2. On session start, Codex fires `SessionStart` → calls `signet hook session-start -H codex --codex-json` → Signet returns identity + memories as `hookSpecificOutput.additionalContext` with `suppressOutput: true`, injected into the model's context window without printing the hook payload to the user transcript.
 3. On every user prompt, Codex fires `UserPromptSubmit` → calls `signet hook user-prompt-submit -H codex --codex-json` → Signet returns per-prompt recalled memories as `hookSpecificOutput.additionalContext` with `suppressOutput: true` so the context is model-visible but not printed into the user transcript. This is blocking — Codex waits for the hook before sending to the model.
 4. On session end, Codex fires `Stop` → calls `signet hook session-end -H codex` → Signet extracts memories from the transcript.
-5. The MCP server exposes `memory_store`, `memory_search`, and other memory tools that Codex can invoke directly during sessions.
+5. The MCP server exposes `memory_store`, `memory_search`, `session_search`, and other tools that Codex can invoke directly during sessions.
 
 Codex `SessionStart` hook timeout defaults to 20 seconds: the Signet CLI
 waits up to `SIGNET_SESSION_START_TIMEOUT` (`15000` ms by default) for
@@ -313,6 +314,7 @@ tools (namespaced as `mcp__signet__*`):
 
 - `memory_store` — save a memory
 - `memory_search` — hybrid recall (vector + keyword)
+- `session_search` — search active or completed session transcripts
 - `memory_list` — list recent memories
 - `memory_modify` — update existing memory
 - `memory_forget` — delete a memory
@@ -738,6 +740,7 @@ Hermes lifecycle.
 | Tool | Description |
 |------|-------------|
 | `memory_search` | Hybrid memory search (keyword + semantic + knowledge graph) |
+| `session_search` | Search active or completed session transcripts |
 | `memory_store` | Store a fact/preference/decision with auto entity extraction |
 | `memory_get` | Retrieve a memory by ID |
 | `memory_list` | List memories with optional filters |

docs/HOOKS.md

@@ -79,12 +79,16 @@ Called when a new agent session begins. Returns memories and context formatted f
 {
   "harness": "openclaw",
   "agentId": "optional-agent-id",
+  "harnessAgentId": "optional-harness-native-subagent-id",
+  "parentSessionKey": "optional-parent-session-key",
   "context": "optional context string",
   "sessionKey": "optional-session-identifier"
 }
 ```
 
-`harness` is required. Everything else is optional.
+`harness` is required. Everything else is optional. `agentId` is the Signet
+persistence scope. Harness-native sub-agent identifiers should be sent as
+`harnessAgentId`; they are used only for parent-session inference.
 
 ### Response
 

docs/MCP.md

@@ -51,6 +51,7 @@ When to Use MCP vs Hooks
 | Session start/end lifecycle | Hooks |
 | Automatic memory extraction after each prompt | Hooks |
 | Agent wants to search memories mid-conversation | MCP (`memory_search`) |
+| Sub-agent needs to inspect its parent session | MCP (`session_search`) |
 | Agent wants to store a specific fact | MCP (`memory_store`) |
 | Agent needs to run a command with secrets | MCP (`secret_exec`) |
 | Compaction boundary handling | Hooks |
@@ -302,6 +303,40 @@ whose memory IDs were actually recorded for the given session and agent.
 are supported for backward compatibility, but the MCP tool is recorded
 immediately on the current turn.
 
+### session_search
+
+Search active or completed session transcripts. This is the pull side of
+sub-agent context continuity: session-start injects a compact parent context
+block when Signet can infer the parent, while `session_search` lets the child
+query the parent transcript on demand.
+
+**Parameters:**
+
+| Name | Type | Required | Description |
+|------|------|----------|-------------|
+| `query` | string | yes | Transcript search query |
+| `session_key` | string | no | Specific transcript session key to search |
+| `current_session_key` | string | no | Current session key; OpenClaw lineage can resolve this to the parent |
+| `agent_id` | string | no | Agent scope (default `default`) |
+| `project` | string | no | Optional project path filter |
+| `limit` | number | no | Max hits to return (default 10, max 20) |
+
+**Returns:** Object with `query`, `hits`, and `count`. Each hit includes
+`sessionKey`, `project`, `updatedAt`, `excerpt`, and `rank`.
+
+**Example:**
+
+```json
+{
+  "query": "trunk ports",
+  "current_session_key": "agent:nicholai:subagent:abc123",
+  "agent_id": "nicholai",
+  "limit": 5
+}
+```
+
+**Daemon endpoint:** `POST /api/sessions/search`
+
 ### agent_peers
 
 List currently active peer sessions for cross-agent coordination.

docs/specs/INDEX.md

@@ -751,7 +751,7 @@ Legend:
 | `marketplace-reviews-cloudflare-staging` | planning | `docs/specs/planning/marketplace-reviews-cloudflare-staging.md` | - | - | |
 | `predictor-agent-feedback` | approved | `docs/specs/approved/predictor-agent-feedback.md` | `predictive-memory-scorer` | - | |
 | `retroactive-supersession` | planning | `docs/specs/planning/retroactive-supersession.md` | `knowledge-architecture-schema` | - | Informed by MSAM-COMPARISON, RESEARCH-COMPETITIVE-SYSTEMS |
-| `sub-agent-context-continuity` | planning | `docs/specs/planning/sub-agent-context-continuity.md` | `session-continuity-protocol`, `multi-agent-support` | - | Parent transcript query + deterministic sub-agent inheritance (issue #315) |
+| `sub-agent-context-continuity` | approved | `docs/specs/approved/sub-agent-context-continuity.md` | `session-continuity-protocol`, `multi-agent-support` | - | Parent transcript query + deterministic sub-agent inheritance (issue #315) |
 | `ontology-evolution-core` | planning | `docs/specs/planning/ontology-evolution-core.md` | `knowledge-architecture-schema`, `desire-paths-epic` | `ontology-governance-workflow` | Confidence/provenance edges, co-occurrence signals, typed relationships, temporal lineage |
 | `ontology-governance-workflow` | planning | `docs/specs/planning/ontology-governance-workflow.md` | `ontology-evolution-core`, `knowledge-architecture-schema` | - | Proposal/review workflow for ontology-impacting schema changes |
 | `ssm-foundation-evaluation` | planning | `docs/specs/planning/ssm-foundation-evaluation.md` | `predictive-memory-scorer`, `desire-paths-epic` | `ssm-temporal-backbone` | Benchmark harness and canary gates for SSM adoption |