diff --git a/packages/ai/cohere/src/index.ts b/packages/ai/cohere/src/index.ts
index 6d8a13b0..7740d0f7 100644
--- a/packages/ai/cohere/src/index.ts
+++ b/packages/ai/cohere/src/index.ts
@@ -4,27 +4,65 @@ interface Config {
   baseUrl?: string;
 }
 
+const DEFAULT_BASE = "https://api.cohere.ai/compatibility";
+
+function chatCompletionsUrl(baseUrl: string): string {
+  const base = baseUrl.replace(/\/+$/, '');
+  return base.endsWith('/v1') ? `${base}/chat/completions` : `${base}/v1/chat/completions`;
+}
+
 export default defineAi<Config>({
-  id: 'ai-cohere',
-  label: 'Cohere',
-  defaultModel: 'command-r-plus',
-  models: ['command-r-plus'],
-
-  async generate(ctx, prompt, _opts, _config) {
-    const apiKey = ctx.secret('COHERE_API_KEY');
-    if (!apiKey) throw new Error('COHERE_API_KEY not in vault — run `sh1pt promote ai setup`');
-    ctx.log(`[stub] ai-cohere · ${prompt.length} chars in — integration pending`);
-    return { text: '[stub — ai-cohere integration not yet implemented]', model: 'command-r-plus' };
+  id: "ai-cohere",
+  label: "Cohere",
+  defaultModel: "command-a-plus-05-2026",
+  models: ["command-a-plus-05-2026", "command-a-03-2025", "command-r-plus", "command-r"],
+
+  async generate(ctx, prompt, opts, config) {
+    const apiKey = ctx.secret("COHERE_API_KEY");
+    if (!apiKey) throw new Error('COHERE_API_KEY not in vault - run `sh1pt promote ai setup`');
+    const model = opts.model ?? "command-a-plus-05-2026";
+    ctx.log(`ai-cohere · model=${model} · ${prompt.length} chars in`);
+    if (ctx.dryRun) return { text: '[dry-run]', model };
+
+    const messages: Array<{ role: string; content: string }> = [];
+    if (opts.system) messages.push({ role: 'system', content: opts.system });
+    messages.push({ role: 'user', content: prompt });
+
+    const res = await fetch(chatCompletionsUrl(config.baseUrl ?? DEFAULT_BASE), {
+      method: 'POST',
+      headers: {
+        authorization: `Bearer ${apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model,
+        messages,
+        ...(opts.maxTokens !== undefined ? { max_tokens: opts.maxTokens } : {}),
+        ...(opts.temperature !== undefined ? { temperature: opts.temperature } : {}),
+        ...opts.extra,
+      }),
+    });
+    if (!res.ok) throw new Error(`Cohere ${res.status}: ${(await res.text()).slice(0, 200)}`);
+    const data = (await res.json()) as {
+      choices: Array<{ message?: { content?: string } }>;
+      model?: string;
+      usage?: { prompt_tokens?: number; completion_tokens?: number };
+    };
+    return {
+      text: data.choices[0]?.message?.content ?? '',
+      model: data.model ?? model,
+      inputTokens: data.usage?.prompt_tokens,
+      outputTokens: data.usage?.completion_tokens,
+    };
   },
 
   setup: tokenSetup<Config>({
-    secretKey: 'COHERE_API_KEY',
-    label: 'Cohere',
-    vendorDocUrl: 'https://dashboard.cohere.com',
-    steps: [
-      'Sign in at https://dashboard.cohere.com and create an API key',
-      'Copy the key — usually shown once',
-      'Paste below; sh1pt encrypts it in the vault',
+    secretKey: "COHERE_API_KEY",
+    label: "Cohere",
+    vendorDocUrl: "https://docs.cohere.com/docs/compatibility-api",
+    steps: ["Sign in at https://dashboard.cohere.com and create an API key", "Copy the key - usually shown once", "Paste below; sh1pt encrypts it in the vault"],
+    fields: [
+      { key: 'baseUrl', message: 'OpenAI-compatible base URL (optional; leave blank for the default):' },
     ],
   }),
 });
diff --git a/packages/ai/kimi/src/index.ts b/packages/ai/kimi/src/index.ts
index 6bbf7fd3..4add1398 100644
--- a/packages/ai/kimi/src/index.ts
+++ b/packages/ai/kimi/src/index.ts
@@ -4,27 +4,65 @@ interface Config {
   baseUrl?: string;
 }
 
+const DEFAULT_BASE = "https://api.moonshot.ai";
+
+function chatCompletionsUrl(baseUrl: string): string {
+  const base = baseUrl.replace(/\/+$/, '');
+  return base.endsWith('/v1') ? `${base}/chat/completions` : `${base}/v1/chat/completions`;
+}
+
 export default defineAi<Config>({
-  id: 'ai-kimi',
-  label: 'Kimi (Moonshot)',
-  defaultModel: 'kimi-k2-0905-preview',
-  models: ['kimi-k2-0905-preview'],
-
-  async generate(ctx, prompt, _opts, _config) {
-    const apiKey = ctx.secret('MOONSHOT_API_KEY');
-    if (!apiKey) throw new Error('MOONSHOT_API_KEY not in vault — run `sh1pt promote ai setup`');
-    ctx.log(`[stub] ai-kimi · ${prompt.length} chars in — integration pending`);
-    return { text: '[stub — ai-kimi integration not yet implemented]', model: 'kimi-k2-0905-preview' };
+  id: "ai-kimi",
+  label: "Kimi (Moonshot)",
+  defaultModel: "kimi-k2.6",
+  models: ["kimi-k2.6", "kimi-k2-0905-preview"],
+
+  async generate(ctx, prompt, opts, config) {
+    const apiKey = ctx.secret("MOONSHOT_API_KEY");
+    if (!apiKey) throw new Error('MOONSHOT_API_KEY not in vault - run `sh1pt promote ai setup`');
+    const model = opts.model ?? "kimi-k2.6";
+    ctx.log(`ai-kimi · model=${model} · ${prompt.length} chars in`);
+    if (ctx.dryRun) return { text: '[dry-run]', model };
+
+    const messages: Array<{ role: string; content: string }> = [];
+    if (opts.system) messages.push({ role: 'system', content: opts.system });
+    messages.push({ role: 'user', content: prompt });
+
+    const res = await fetch(chatCompletionsUrl(config.baseUrl ?? DEFAULT_BASE), {
+      method: 'POST',
+      headers: {
+        authorization: `Bearer ${apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model,
+        messages,
+        ...(opts.maxTokens !== undefined ? { max_tokens: opts.maxTokens } : {}),
+        ...(opts.temperature !== undefined ? { temperature: opts.temperature } : {}),
+        ...opts.extra,
+      }),
+    });
+    if (!res.ok) throw new Error(`Kimi (Moonshot) ${res.status}: ${(await res.text()).slice(0, 200)}`);
+    const data = (await res.json()) as {
+      choices: Array<{ message?: { content?: string } }>;
+      model?: string;
+      usage?: { prompt_tokens?: number; completion_tokens?: number };
+    };
+    return {
+      text: data.choices[0]?.message?.content ?? '',
+      model: data.model ?? model,
+      inputTokens: data.usage?.prompt_tokens,
+      outputTokens: data.usage?.completion_tokens,
+    };
   },
 
   setup: tokenSetup<Config>({
-    secretKey: 'MOONSHOT_API_KEY',
-    label: 'Kimi (Moonshot)',
-    vendorDocUrl: 'https://platform.moonshot.ai',
-    steps: [
-      'Sign in at https://platform.moonshot.ai and create an API key',
-      'Copy the key — usually shown once',
-      'Paste below; sh1pt encrypts it in the vault',
+    secretKey: "MOONSHOT_API_KEY",
+    label: "Kimi (Moonshot)",
+    vendorDocUrl: "https://platform.kimi.ai/docs/api/overview",
+    steps: ["Sign in at https://platform.kimi.ai and create an API key", "Copy the key - usually shown once", "Paste below; sh1pt encrypts it in the vault"],
+    fields: [
+      { key: 'baseUrl', message: 'OpenAI-compatible base URL (optional; leave blank for the default):' },
     ],
   }),
 });
diff --git a/packages/ai/novita/src/index.ts b/packages/ai/novita/src/index.ts
index 70f3b05a..66a17cad 100644
--- a/packages/ai/novita/src/index.ts
+++ b/packages/ai/novita/src/index.ts
@@ -4,27 +4,65 @@ interface Config {
   baseUrl?: string;
 }
 
+const DEFAULT_BASE = "https://api.novita.ai/openai";
+
+function chatCompletionsUrl(baseUrl: string): string {
+  const base = baseUrl.replace(/\/+$/, '');
+  return base.endsWith('/v1') ? `${base}/chat/completions` : `${base}/v1/chat/completions`;
+}
+
 export default defineAi<Config>({
-  id: 'ai-novita',
-  label: 'NovitaAI',
-  defaultModel: 'meta-llama/llama-3.3-70b-instruct',
-  models: ['meta-llama/llama-3.3-70b-instruct'],
-
-  async generate(ctx, prompt, _opts, _config) {
-    const apiKey = ctx.secret('NOVITA_API_KEY');
-    if (!apiKey) throw new Error('NOVITA_API_KEY not in vault — run `sh1pt promote ai setup`');
-    ctx.log(`[stub] ai-novita · ${prompt.length} chars in — integration pending`);
-    return { text: '[stub — ai-novita integration not yet implemented]', model: 'meta-llama/llama-3.3-70b-instruct' };
+  id: "ai-novita",
+  label: "NovitaAI",
+  defaultModel: "meta-llama/llama-3.3-70b-instruct",
+  models: ["meta-llama/llama-3.3-70b-instruct", "deepseek/deepseek-v3-0324", "qwen/qwen3-235b-a22b-fp8"],
+
+  async generate(ctx, prompt, opts, config) {
+    const apiKey = ctx.secret("NOVITA_API_KEY");
+    if (!apiKey) throw new Error('NOVITA_API_KEY not in vault - run `sh1pt promote ai setup`');
+    const model = opts.model ?? "meta-llama/llama-3.3-70b-instruct";
+    ctx.log(`ai-novita · model=${model} · ${prompt.length} chars in`);
+    if (ctx.dryRun) return { text: '[dry-run]', model };
+
+    const messages: Array<{ role: string; content: string }> = [];
+    if (opts.system) messages.push({ role: 'system', content: opts.system });
+    messages.push({ role: 'user', content: prompt });
+
+    const res = await fetch(chatCompletionsUrl(config.baseUrl ?? DEFAULT_BASE), {
+      method: 'POST',
+      headers: {
+        authorization: `Bearer ${apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model,
+        messages,
+        ...(opts.maxTokens !== undefined ? { max_tokens: opts.maxTokens } : {}),
+        ...(opts.temperature !== undefined ? { temperature: opts.temperature } : {}),
+        ...opts.extra,
+      }),
+    });
+    if (!res.ok) throw new Error(`NovitaAI ${res.status}: ${(await res.text()).slice(0, 200)}`);
+    const data = (await res.json()) as {
+      choices: Array<{ message?: { content?: string } }>;
+      model?: string;
+      usage?: { prompt_tokens?: number; completion_tokens?: number };
+    };
+    return {
+      text: data.choices[0]?.message?.content ?? '',
+      model: data.model ?? model,
+      inputTokens: data.usage?.prompt_tokens,
+      outputTokens: data.usage?.completion_tokens,
+    };
   },
 
   setup: tokenSetup<Config>({
-    secretKey: 'NOVITA_API_KEY',
-    label: 'NovitaAI',
-    vendorDocUrl: 'https://novita.ai',
-    steps: [
-      'Sign in at https://novita.ai and create an API key',
-      'Copy the key — usually shown once',
-      'Paste below; sh1pt encrypts it in the vault',
+    secretKey: "NOVITA_API_KEY",
+    label: "NovitaAI",
+    vendorDocUrl: "https://novita.ai/docs/api-reference/model-apis-llm-create-chat-completion",
+    steps: ["Sign in at https://novita.ai and create an API key", "Copy the key - usually shown once", "Paste below; sh1pt encrypts it in the vault"],
+    fields: [
+      { key: 'baseUrl', message: 'OpenAI-compatible base URL (optional; leave blank for the default):' },
     ],
   }),
 });
diff --git a/packages/ai/parasail/src/index.ts b/packages/ai/parasail/src/index.ts
index 88d4be22..d17f5a2c 100644
--- a/packages/ai/parasail/src/index.ts
+++ b/packages/ai/parasail/src/index.ts
@@ -4,27 +4,65 @@ interface Config {
   baseUrl?: string;
 }
 
+const DEFAULT_BASE = "https://api.parasail.io";
+
+function chatCompletionsUrl(baseUrl: string): string {
+  const base = baseUrl.replace(/\/+$/, '');
+  return base.endsWith('/v1') ? `${base}/chat/completions` : `${base}/v1/chat/completions`;
+}
+
 export default defineAi<Config>({
-  id: 'ai-parasail',
-  label: 'Parasail',
-  defaultModel: 'PARASAIL_API_KEY',
-  models: ['PARASAIL_API_KEY'],
-
-  async generate(ctx, prompt, _opts, _config) {
-    const apiKey = ctx.secret('https://parasail.io');
-    if (!apiKey) throw new Error('https://parasail.io not in vault — run `sh1pt promote ai setup`');
-    ctx.log(`[stub] ai-parasail · ${prompt.length} chars in — integration pending`);
-    return { text: '[stub — ai-parasail integration not yet implemented]', model: 'PARASAIL_API_KEY' };
+  id: "ai-parasail",
+  label: "Parasail",
+  defaultModel: "parasail-llama-33-70b-fp8",
+  models: ["parasail-llama-33-70b-fp8", "parasail-deepseek-r1", "parasail-qwen3-32b"],
+
+  async generate(ctx, prompt, opts, config) {
+    const apiKey = ctx.secret("PARASAIL_API_KEY");
+    if (!apiKey) throw new Error('PARASAIL_API_KEY not in vault - run `sh1pt promote ai setup`');
+    const model = opts.model ?? "parasail-llama-33-70b-fp8";
+    ctx.log(`ai-parasail · model=${model} · ${prompt.length} chars in`);
+    if (ctx.dryRun) return { text: '[dry-run]', model };
+
+    const messages: Array<{ role: string; content: string }> = [];
+    if (opts.system) messages.push({ role: 'system', content: opts.system });
+    messages.push({ role: 'user', content: prompt });
+
+    const res = await fetch(chatCompletionsUrl(config.baseUrl ?? DEFAULT_BASE), {
+      method: 'POST',
+      headers: {
+        authorization: `Bearer ${apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model,
+        messages,
+        ...(opts.maxTokens !== undefined ? { max_completion_tokens: opts.maxTokens } : {}),
+        ...(opts.temperature !== undefined ? { temperature: opts.temperature } : {}),
+        ...opts.extra,
+      }),
+    });
+    if (!res.ok) throw new Error(`Parasail ${res.status}: ${(await res.text()).slice(0, 200)}`);
+    const data = (await res.json()) as {
+      choices: Array<{ message?: { content?: string } }>;
+      model?: string;
+      usage?: { prompt_tokens?: number; completion_tokens?: number };
+    };
+    return {
+      text: data.choices[0]?.message?.content ?? '',
+      model: data.model ?? model,
+      inputTokens: data.usage?.prompt_tokens,
+      outputTokens: data.usage?.completion_tokens,
+    };
   },
 
   setup: tokenSetup<Config>({
-    secretKey: 'https://parasail.io',
-    label: 'Parasail',
-    vendorDocUrl: '',
-    steps: [
-      'Sign in at  and create an API key',
-      'Copy the key — usually shown once',
-      'Paste below; sh1pt encrypts it in the vault',
+    secretKey: "PARASAIL_API_KEY",
+    label: "Parasail",
+    vendorDocUrl: "https://docs.parasail.io/parasail-docs/cookbooks/chat-completions",
+    steps: ["Sign in at https://parasail.io and create an API key", "Copy the key - usually shown once", "Paste below; sh1pt encrypts it in the vault"],
+    fields: [
+      { key: 'baseUrl', message: 'OpenAI-compatible base URL (optional; leave blank for the default):' },
     ],
   }),
 });
diff --git a/packages/ai/venice/src/index.ts b/packages/ai/venice/src/index.ts
index a0aa321f..5e60fdd1 100644
--- a/packages/ai/venice/src/index.ts
+++ b/packages/ai/venice/src/index.ts
@@ -4,27 +4,65 @@ interface Config {
   baseUrl?: string;
 }
 
+const DEFAULT_BASE = "https://api.venice.ai/api";
+
+function chatCompletionsUrl(baseUrl: string): string {
+  const base = baseUrl.replace(/\/+$/, '');
+  return base.endsWith('/v1') ? `${base}/chat/completions` : `${base}/v1/chat/completions`;
+}
+
 export default defineAi<Config>({
-  id: 'ai-venice',
-  label: 'Venice AI',
-  defaultModel: 'llama-3.3-70b',
-  models: ['llama-3.3-70b'],
-
-  async generate(ctx, prompt, _opts, _config) {
-    const apiKey = ctx.secret('VENICE_API_KEY');
-    if (!apiKey) throw new Error('VENICE_API_KEY not in vault — run `sh1pt promote ai setup`');
-    ctx.log(`[stub] ai-venice · ${prompt.length} chars in — integration pending`);
-    return { text: '[stub — ai-venice integration not yet implemented]', model: 'llama-3.3-70b' };
+  id: "ai-venice",
+  label: "Venice AI",
+  defaultModel: "llama-3.3-70b",
+  models: ["llama-3.3-70b", "qwen3-235b", "dolphin-2.9.2-qwen2-72b"],
+
+  async generate(ctx, prompt, opts, config) {
+    const apiKey = ctx.secret("VENICE_API_KEY");
+    if (!apiKey) throw new Error('VENICE_API_KEY not in vault - run `sh1pt promote ai setup`');
+    const model = opts.model ?? "llama-3.3-70b";
+    ctx.log(`ai-venice · model=${model} · ${prompt.length} chars in`);
+    if (ctx.dryRun) return { text: '[dry-run]', model };
+
+    const messages: Array<{ role: string; content: string }> = [];
+    if (opts.system) messages.push({ role: 'system', content: opts.system });
+    messages.push({ role: 'user', content: prompt });
+
+    const res = await fetch(chatCompletionsUrl(config.baseUrl ?? DEFAULT_BASE), {
+      method: 'POST',
+      headers: {
+        authorization: `Bearer ${apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model,
+        messages,
+        ...(opts.maxTokens !== undefined ? { max_tokens: opts.maxTokens } : {}),
+        ...(opts.temperature !== undefined ? { temperature: opts.temperature } : {}),
+        ...opts.extra,
+      }),
+    });
+    if (!res.ok) throw new Error(`Venice AI ${res.status}: ${(await res.text()).slice(0, 200)}`);
+    const data = (await res.json()) as {
+      choices: Array<{ message?: { content?: string } }>;
+      model?: string;
+      usage?: { prompt_tokens?: number; completion_tokens?: number };
+    };
+    return {
+      text: data.choices[0]?.message?.content ?? '',
+      model: data.model ?? model,
+      inputTokens: data.usage?.prompt_tokens,
+      outputTokens: data.usage?.completion_tokens,
+    };
   },
 
   setup: tokenSetup<Config>({
-    secretKey: 'VENICE_API_KEY',
-    label: 'Venice AI',
-    vendorDocUrl: 'https://venice.ai',
-    steps: [
-      'Sign in at https://venice.ai and create an API key',
-      'Copy the key — usually shown once',
-      'Paste below; sh1pt encrypts it in the vault',
+    secretKey: "VENICE_API_KEY",
+    label: "Venice AI",
+    vendorDocUrl: "https://docs.venice.ai/api-reference/api-spec",
+    steps: ["Sign in at https://venice.ai and create an API key", "Copy the key - usually shown once", "Paste below; sh1pt encrypts it in the vault"],
+    fields: [
+      { key: 'baseUrl', message: 'OpenAI-compatible base URL (optional; leave blank for the default):' },
     ],
   }),
 });
diff --git a/packages/core/src/setup-helpers.ts b/packages/core/src/setup-helpers.ts
index 09aea2b4..d53a55b6 100644
--- a/packages/core/src/setup-helpers.ts
+++ b/packages/core/src/setup-helpers.ts
@@ -196,6 +196,8 @@ export function tokenSetup<C = unknown>(opts: TokenSetupOpts<C>): SetupFn<C> {
       if (val) {
         if (field.secret) await ctx.setSecret(field.key, val);
         else configExtras[field.key] = val;
+      } else if (!field.secret) {
+        configExtras[field.key] = '';
       }
     }
 
diff --git a/packages/core/src/setup.ts b/packages/core/src/setup.ts
index 9a465491..ecb04967 100644
--- a/packages/core/src/setup.ts
+++ b/packages/core/src/setup.ts
@@ -16,7 +16,7 @@
 //
 // Secrets go to the vault via ctx.setSecret — never into config.json.
 
-import { configPath, setAdapterConfig } from './config-store.js';
+import { configPath, getAdapterConfig, setAdapterConfig } from './config-store.js';
 
 export interface SetupPromptDef<T = string> {
   type: 'text' | 'password' | 'select' | 'confirm';
@@ -78,7 +78,9 @@ export async function runSetup<Config = unknown>(
     };
   }
 
-  await setAdapterConfig(adapter.id, result.config);
+  const existingConfig = await getAdapterConfig(adapter.id);
+  const configToSave = mergeAdapterConfig(existingConfig, result.config);
+  await setAdapterConfig(adapter.id, configToSave);
   ctx.log(`  saved → ${configPath()} · adapters.${adapter.id}`);
 
   if (!result.ok && result.manual && result.manual.length > 0) {
@@ -91,3 +93,18 @@ export async function runSetup<Config = unknown>(
 
   return result;
 }
+
+
+function isPlainRecord(value: unknown): value is Record<string, unknown> {
+  return value !== null && typeof value === 'object' && !Array.isArray(value);
+}
+
+function mergeAdapterConfig(existing: unknown, next: unknown): unknown {
+  if (!isPlainRecord(existing) || !isPlainRecord(next)) return next;
+  const merged: Record<string, unknown> = { ...existing };
+  for (const [key, value] of Object.entries(next)) {
+    if (value === '') delete merged[key];
+    else merged[key] = value;
+  }
+  return merged;
+}