diff --git a/packages/local-runtime/src/context/token-counter-adapters/openai-chat-tokenizers.ts b/packages/local-runtime/src/context/token-counter-adapters/openai-chat-tokenizers.ts
index 496230d3e..97295d760 100644
--- a/packages/local-runtime/src/context/token-counter-adapters/openai-chat-tokenizers.ts
+++ b/packages/local-runtime/src/context/token-counter-adapters/openai-chat-tokenizers.ts
@@ -16,6 +16,9 @@ const ZHIPU_PATHS = new Set([
   '/api/paas/v4',
   '/api/paas/v4/chat/completions',
   '/api/paas/v4/tokenizer',
+  '/api/coding/paas/v4',
+  '/api/coding/paas/v4/chat/completions',
+  '/api/coding/paas/v4/tokenizer',
 ]);
 const KIMI_PATHS = new Set(['/v1', '/v1/chat/completions', '/v1/tokenizers/estimate-token-count']);
 
@@ -73,7 +76,7 @@ export const zhipuTokenizerTokenCounterAdapter: RemoteTokenCounterAdapter = {
   buildRequest: (ctx) =>
     buildOpenAIChatTokenizerRequest(
       ctx,
-      '/api/paas/v4/tokenizer',
+      zhipuTokenizerEndpointPath(ctx.model.baseUrl),
       'zhipu-tokenizer',
       ZHIPU_CHAT_COMPAT,
     ),
@@ -94,6 +97,20 @@ export const kimiEstimateTokenCountAdapter: RemoteTokenCounterAdapter = {
   parseTokens: parseDataTotalTokens,
 };
 
+/**
+ * Zhipu exposes the same surface under two plan roots: `/api/paas/v4` (general
+ * API plan) and `/api/coding/paas/v4` (coding plan). The tokenizer endpoint
+ * lives under the same root as the configured base URL, so keep the request on
+ * the caller's plan instead of hardcoding the general-plan path.
+ */
+function zhipuTokenizerEndpointPath(baseUrl: string): string {
+  const url = parseUrl(baseUrl);
+  const pathname = url ? normalizePathname(url.pathname) : '';
+  return pathname === '/api/coding/paas/v4' || pathname.startsWith('/api/coding/paas/v4/')
+    ? '/api/coding/paas/v4/tokenizer'
+    : '/api/paas/v4/tokenizer';
+}
+
 function matchesOfficialOpenAIChatProvider(
   ctx: RemoteTokenCountContext,
   allowedHosts: ReadonlySet<string>,
diff --git a/packages/local-runtime/src/context/token-counter-adapters/responses.ts b/packages/local-runtime/src/context/token-counter-adapters/responses.ts
index 602494fba..02889caa9 100644
--- a/packages/local-runtime/src/context/token-counter-adapters/responses.ts
+++ b/packages/local-runtime/src/context/token-counter-adapters/responses.ts
@@ -91,7 +91,7 @@ export function buildResponsesInputTokensUrl(
     if (options.stripMessagesCompatibilityPrefix) {
       rootPath = stripMessagesCompatibilityPrefix(rootPath);
     }
-    if (!rootPath.endsWith('/v1')) {
+    if (!endsWithVersionSegment(rootPath)) {
       rootPath = `${rootPath}/v1`;
     }
     url.pathname = `${rootPath}/responses/input_tokens`.replace(/\/{2,}/gu, '/');
@@ -109,11 +109,21 @@ export function buildResponsesInputTokensUrl(
     if (options.stripMessagesCompatibilityPrefix) {
       root = stripMessagesCompatibilityPrefix(root);
     }
-    if (!root.endsWith('/v1')) root = `${root}/v1`;
+    if (!endsWithVersionSegment(root)) root = `${root}/v1`;
     return `${root}/responses/input_tokens`;
   }
 }
 
+/**
+ * Base URLs of OpenAI-compatible gateways are often already versioned
+ * (`/v1`, but also `/api/paas/v4`, `/api/coding/paas/v4`, ...). Appending
+ * `/v1` after an existing version segment can only produce a path that no
+ * gateway serves, so treat any trailing `/v<digits>` as the API root.
+ */
+function endsWithVersionSegment(root: string): boolean {
+  return /\/v\d+$/u.test(root);
+}
+
 function stripMessagesCompatibilityPrefix(root: string): string {
   return root.replace(/\/anthropic(?=\/v1$|$)/u, '');
 }
diff --git a/packages/local-runtime/test/unit/token-counter-adapter-routing.test.ts b/packages/local-runtime/test/unit/token-counter-adapter-routing.test.ts
new file mode 100644
index 000000000..4cb8fdc89
--- /dev/null
+++ b/packages/local-runtime/test/unit/token-counter-adapter-routing.test.ts
@@ -0,0 +1,106 @@
+import { describe, expect, it } from 'vitest';
+
+import { resolveRemoteTokenCounterAdapter } from '../../src/context/token-counter-adapters/registry.js';
+import { buildResponsesInputTokensUrl } from '../../src/context/token-counter-adapters/responses.js';
+import type { RemoteTokenCountContext } from '../../src/context/token-counter-adapters/types.js';
+
+function makeContext(baseUrl: string, api = 'openai-completions'): RemoteTokenCountContext {
+  return {
+    model: { id: 'glm-5.3', api, provider: 'custom_provider', baseUrl, input: ['text'] },
+    apiKey: 'sk-test',
+    systemPrompt: 'hi',
+    messages: [],
+    tools: [],
+  } as unknown as RemoteTokenCountContext;
+}
+
+describe('remote token counter adapter routing (issue #258)', () => {
+  it('routes Zhipu general API-plan base URLs to the zhipu tokenizer', () => {
+    for (const baseUrl of [
+      'https://api.z.ai/api/paas/v4',
+      'https://open.bigmodel.cn/api/paas/v4/chat/completions',
+    ]) {
+      const ctx = makeContext(baseUrl);
+      const adapter = resolveRemoteTokenCounterAdapter(ctx);
+      expect(adapter?.id).toBe('zhipu-tokenizer');
+      expect(adapter?.buildRequest(ctx)?.url).toMatch(/\/api\/paas\/v4\/tokenizer$/u);
+    }
+  });
+
+  it('routes Zhipu coding-plan base URLs to the zhipu tokenizer on the coding-plan root', () => {
+    for (const host of ['api.z.ai', 'open.bigmodel.cn']) {
+      for (const path of ['/api/coding/paas/v4', '/api/coding/paas/v4/chat/completions']) {
+        const ctx = makeContext(`https://${host}${path}`);
+        const adapter = resolveRemoteTokenCounterAdapter(ctx);
+        expect(adapter?.id).toBe('zhipu-tokenizer');
+        expect(adapter?.buildRequest(ctx)?.url).toBe(
+          `https://${host}/api/coding/paas/v4/tokenizer`,
+        );
+      }
+    }
+  });
+
+  it('routes Kimi base URLs to the kimi estimator', () => {
+    const adapter = resolveRemoteTokenCounterAdapter(makeContext('https://api.moonshot.cn/v1'));
+    expect(adapter?.id).toBe('kimi-estimate-token-count');
+  });
+
+  it('keeps unrelated OpenAI-compatible providers on generic-responses', () => {
+    const adapter = resolveRemoteTokenCounterAdapter(
+      makeContext('https://gateway.example.com/v1'),
+    );
+    expect(adapter?.id).toBe('generic-responses');
+  });
+});
+
+describe('buildResponsesInputTokensUrl version handling (issue #258)', () => {
+  it('keeps /v1 base URLs unchanged', () => {
+    expect(buildResponsesInputTokensUrl('https://api.example.com/v1')).toBe(
+      'https://api.example.com/v1/responses/input_tokens',
+    );
+  });
+
+  it('appends /v1 when the base URL has no version segment', () => {
+    expect(buildResponsesInputTokensUrl('https://api.example.com')).toBe(
+      'https://api.example.com/v1/responses/input_tokens',
+    );
+    expect(buildResponsesInputTokensUrl('https://api.example.com/openai')).toBe(
+      'https://api.example.com/openai/v1/responses/input_tokens',
+    );
+  });
+
+  it('does not double the version for base URLs already ending in a version segment', () => {
+    expect(buildResponsesInputTokensUrl('https://api.z.ai/api/coding/paas/v4')).toBe(
+      'https://api.z.ai/api/coding/paas/v4/responses/input_tokens',
+    );
+    expect(buildResponsesInputTokensUrl('https://gateway.example.com/v2')).toBe(
+      'https://gateway.example.com/v2/responses/input_tokens',
+    );
+  });
+
+  it('strips chat/completions and responses suffixes before resolving the root', () => {
+    expect(buildResponsesInputTokensUrl('https://api.example.com/v1/chat/completions')).toBe(
+      'https://api.example.com/v1/responses/input_tokens',
+    );
+    expect(
+      buildResponsesInputTokensUrl('https://open.bigmodel.cn/api/paas/v4/chat/completions'),
+    ).toBe('https://open.bigmodel.cn/api/paas/v4/responses/input_tokens');
+    expect(buildResponsesInputTokensUrl('https://api.example.com/v1/responses')).toBe(
+      'https://api.example.com/v1/responses/input_tokens',
+    );
+  });
+
+  it('keeps the MiniMax anthropic-compat stripping behavior stable', () => {
+    expect(
+      buildResponsesInputTokensUrl('https://api.minimaxi.com/anthropic', {
+        stripMessagesCompatibilityPrefix: true,
+      }),
+    ).toBe('https://api.minimaxi.com/v1/responses/input_tokens');
+  });
+
+  it('falls back to string handling for non-URL base values', () => {
+    expect(buildResponsesInputTokensUrl('not a url/api/paas/v4')).toBe(
+      'not a url/api/paas/v4/responses/input_tokens',
+    );
+  });
+});
diff --git a/release/public-source.json b/release/public-source.json
index 2f0c50fcb..d4b06b932 100644
--- a/release/public-source.json
+++ b/release/public-source.json
@@ -2653,6 +2653,7 @@
     "packages/local-runtime/test/unit/thread-goal/host-integration-settlement.test.ts",
     "packages/local-runtime/test/unit/thread-goal/host-integration-testkit.ts",
     "packages/local-runtime/test/unit/thread-goal/host-integration-verifier-charge-rearm.test.ts",
+    "packages/local-runtime/test/unit/token-counter-adapter-routing.test.ts",
     "packages/mcode-tools-host/README.md",
     "packages/mcode-tools-host/package.json",
     "packages/mcode-tools-host/src/contracts.ts",
diff --git a/test/vitest-suites.json b/test/vitest-suites.json
index 2063f581a..94c5ab43b 100644
--- a/test/vitest-suites.json
+++ b/test/vitest-suites.json
@@ -149,7 +149,8 @@
       "packages/tui/test/unit/tui-feature-viewport-contract.test.ts",
       "packages/tui/test/unit/tui-session-history-flow.test.ts",
       "packages/tui/test/unit/tui-session-mutation-panels.test.ts",
-      "packages/tui/test/unit/tui-surface-host.test.ts"
+      "packages/tui/test/unit/tui-surface-host.test.ts",
+      "packages/local-runtime/test/unit/token-counter-adapter-routing.test.ts"
     ],
     "status-contract": [
       "packages/tui/test/unit/tui-build-mode-contract.test.ts"