fix: persist embedded runtime context budget

CHANGELOG.md

@@ -14,6 +14,7 @@ Docs: https://docs.openclaw.ai
 - Slack/files: return non-image `download-file` results as local file paths instead of image payloads, and include Slack file IDs in inbound file placeholders so agents can call `download-file`. Fixes #71212. Thanks @teamrazo.
 - Browser control: scope standalone loopback auth to the resolved active gateway credential and fail closed when password mode lacks a resolved password, so inactive tokens or passwords no longer authorize browser routes. Fixes #65626. (#65639) Thanks @coygeek.
 - Control UI/Codex harness: emit native Codex app-server assistant and lifecycle completion events so live webchat runs stop spinning without needing a transcript reload fallback. (#70815) Thanks @lesaai.
+- Agents/sessions: persist the runtime-resolved context budget from embedded agent runs, so Codex GPT-5.5 sessions keep the catalog/runtime context cap instead of falling back to the generic 200k status value. Fixes #71294. Thanks @tud0r.
 - Discord/replies: run `message_sending` plugin hooks for Discord reply delivery, including DM targets, so plugins can transform or cancel outbound Discord replies consistently with other channels. Fixes #59350. (#71094) Thanks @wei840222.
 - Control UI/commands: carry provider-owned thinking option ids/labels in session rows and defaults so fresh sessions show and accept dynamic modes such as `adaptive`, `xhigh`, and `max`. Fixes #71269. Thanks @Young-Khalil.
 - Image generation: make explicit `model=` overrides exact-only so failed `openai/gpt-image-2` requests no longer fall through to Gemini or other configured providers, and update `image_generate list` to mention OpenAI Codex OAuth as valid auth for `openai/gpt-image-2`. Fixes #71290 and #71231. Thanks @Young-Khalil and @steipete.

src/agents/command/session-store.test.ts

@@ -185,6 +185,47 @@ describe("updateSessionStoreAfterAgentRun", () => {
     });
   });
 
+  it("uses the runtime context budget from agent metadata instead of cold fallback", async () => {
+    await withTempSessionStore(async ({ storePath }) => {
+      const cfg = {} as OpenClawConfig;
+      const sessionKey = "agent:main:explicit:test-runtime-context";
+      const sessionId = "test-runtime-context-session";
+      const sessionStore: Record<string, SessionEntry> = {
+        [sessionKey]: {
+          sessionId,
+          updatedAt: 1,
+        },
+      };
+      await fs.writeFile(storePath, JSON.stringify(sessionStore, null, 2));
+
+      const result: EmbeddedPiRunResult = {
+        meta: {
+          durationMs: 1,
+          agentMeta: {
+            sessionId,
+            provider: "openai-codex",
+            model: "gpt-5.5",
+            contextTokens: 400_000,
+          },
+        },
+      };
+
+      await updateSessionStoreAfterAgentRun({
+        cfg,
+        sessionId,
+        sessionKey,
+        storePath,
+        sessionStore,
+        defaultProvider: "openai-codex",
+        defaultModel: "gpt-5.5",
+        result,
+      });
+
+      expect(sessionStore[sessionKey]?.contextTokens).toBe(400_000);
+      expect(loadSessionStore(storePath)[sessionKey]?.contextTokens).toBe(400_000);
+    });
+  });
+
   it("clears the embedded harness pin after a CLI run", async () => {
     await withTempSessionStore(async ({ storePath }) => {
       const cfg = {

src/agents/command/session-store.ts

@@ -30,6 +30,13 @@ function resolveNonNegativeNumber(value: number | undefined): number | undefined
   return typeof value === "number" && Number.isFinite(value) && value >= 0 ? value : undefined;
 }
 
+function resolvePositiveInteger(value: number | undefined): number | undefined {
+  if (typeof value !== "number" || !Number.isFinite(value) || value <= 0) {
+    return undefined;
+  }
+  return Math.floor(value);
+}
+
 export async function updateSessionStoreAfterAgentRun(params: {
   cfg: OpenClawConfig;
   contextTokensOverride?: number;
@@ -62,16 +69,19 @@ export async function updateSessionStoreAfterAgentRun(params: {
   const modelUsed = result.meta.agentMeta?.model ?? fallbackModel ?? defaultModel;
   const providerUsed = result.meta.agentMeta?.provider ?? fallbackProvider ?? defaultProvider;
   const agentHarnessId = normalizeOptionalString(result.meta.agentMeta?.agentHarnessId);
+  const runtimeContextTokens = resolvePositiveInteger(result.meta.agentMeta?.contextTokens);
   const contextTokens =
-    typeof params.contextTokensOverride === "number" && params.contextTokensOverride > 0
-      ? params.contextTokensOverride
-      : ((await getContextModule()).resolveContextTokensForModel({
-          cfg,
-          provider: providerUsed,
-          model: modelUsed,
-          fallbackContextTokens: DEFAULT_CONTEXT_TOKENS,
-          allowAsyncLoad: false,
-        }) ?? DEFAULT_CONTEXT_TOKENS);
+    runtimeContextTokens !== undefined
+      ? runtimeContextTokens
+      : typeof params.contextTokensOverride === "number" && params.contextTokensOverride > 0
+        ? params.contextTokensOverride
+        : ((await getContextModule()).resolveContextTokensForModel({
+            cfg,
+            provider: providerUsed,
+            model: modelUsed,
+            fallbackContextTokens: DEFAULT_CONTEXT_TOKENS,
+            allowAsyncLoad: false,
+          }) ?? DEFAULT_CONTEXT_TOKENS);
 
   const entry = sessionStore[sessionKey] ?? {
     sessionId,

src/agents/pi-embedded-runner/run.ts

@@ -777,6 +777,7 @@ export async function runEmbeddedPiAgent(
                 sessionId: params.sessionId,
                 provider,
                 model: model.id,
+                contextTokens: ctxInfo.tokens,
                 usageAccumulator,
                 lastRunPromptUsage,
                 lastTurnTotal,
@@ -1371,6 +1372,7 @@ export async function runEmbeddedPiAgent(
                   sessionId: sessionIdUsed,
                   provider,
                   model: model.id,
+                  contextTokens: ctxInfo.tokens,
                   usageAccumulator,
                   lastRunPromptUsage,
                   lastAssistant: sessionLastAssistant,
@@ -1426,6 +1428,7 @@ export async function runEmbeddedPiAgent(
                     sessionId: sessionIdUsed,
                     provider,
                     model: model.id,
+                    contextTokens: ctxInfo.tokens,
                     usageAccumulator,
                     lastRunPromptUsage,
                     lastAssistant: sessionLastAssistant,
@@ -1465,6 +1468,7 @@ export async function runEmbeddedPiAgent(
                     sessionId: sessionIdUsed,
                     provider,
                     model: model.id,
+                    contextTokens: ctxInfo.tokens,
                     usageAccumulator,
                     lastRunPromptUsage,
                     lastAssistant: sessionLastAssistant,
@@ -1776,6 +1780,7 @@ export async function runEmbeddedPiAgent(
             sessionId: sessionIdUsed,
             provider: sessionLastAssistant?.provider ?? provider,
             model: sessionLastAssistant?.model ?? model.id,
+            contextTokens: ctxInfo.tokens,
             agentHarnessId: attempt.agentHarnessId,
             usage: usageMeta.usage,
             lastCallUsage: usageMeta.lastCallUsage,

src/agents/pi-embedded-runner/run/helpers.ts

@@ -124,6 +124,7 @@ export function buildErrorAgentMeta(params: {
   sessionId: string;
   provider: string;
   model: string;
+  contextTokens?: number;
   usageAccumulator: UsageAccumulator;
   lastRunPromptUsage: UsageSnapshot | undefined;
   lastAssistant?: { usage?: unknown } | null;
@@ -139,6 +140,7 @@ export function buildErrorAgentMeta(params: {
     sessionId: params.sessionId,
     provider: params.provider,
     model: params.model,
+    ...(params.contextTokens ? { contextTokens: params.contextTokens } : {}),
     ...(usageMeta.usage ? { usage: usageMeta.usage } : {}),
     ...(usageMeta.lastCallUsage ? { lastCallUsage: usageMeta.lastCallUsage } : {}),
     ...(usageMeta.promptTokens ? { promptTokens: usageMeta.promptTokens } : {}),

src/agents/pi-embedded-runner/types.ts

@@ -6,6 +6,7 @@ export type EmbeddedPiAgentMeta = {
   sessionId: string;
   provider: string;
   model: string;
+  contextTokens?: number;
   agentHarnessId?: string;
   cliSessionBinding?: CliSessionBinding;
   compactionCount?: number;

src/auto-reply/reply/agent-runner.ts

@@ -1324,7 +1324,14 @@ export async function runReplyAgent(params: {
     const cliSessionBinding = isCliProvider(providerUsed, cfg)
       ? runResult.meta?.agentMeta?.cliSessionBinding
       : undefined;
+    const runtimeContextTokens =
+      typeof runResult.meta?.agentMeta?.contextTokens === "number" &&
+      Number.isFinite(runResult.meta.agentMeta.contextTokens) &&
+      runResult.meta.agentMeta.contextTokens > 0
+        ? Math.floor(runResult.meta.agentMeta.contextTokens)
+        : undefined;
     const contextTokensUsed =
+      runtimeContextTokens ??
       resolveContextTokensForModel({
         cfg,
         provider: providerUsed,
@@ -1332,7 +1339,8 @@ export async function runReplyAgent(params: {
         contextTokensOverride: agentCfgContextTokens,
         fallbackContextTokens: activeSessionEntry?.contextTokens ?? DEFAULT_CONTEXT_TOKENS,
         allowAsyncLoad: false,
-      }) ?? DEFAULT_CONTEXT_TOKENS;
+      }) ??
+      DEFAULT_CONTEXT_TOKENS;
 
     await persistRunSessionUsage({
       storePath,