Fix context window, drop model tiers, and pick a provider before a model

src/agent/agent-search.test.ts

@@ -12,17 +12,14 @@ const fixtures: AgentProfile[] = [
   {
     id: "greybeard",
     description: "Seasoned architect — reviews for design and backwards compatibility",
-    tier: "clever",
   },
   {
     id: "critique",
     description: "Code quality reviewer — tests assumptions and security smells",
-    tier: "standard",
   },
   {
     id: "scout",
     description: "Fast codebase explorer — maps structure and entry points",
-    tier: "fast",
   },
 ];
 
@@ -71,13 +68,11 @@ describe("formatAgentSearchResults", () => {
         id: "draper",
         description: "PR design reviewer from marketplace",
         source: "claude",
-        tier: "standard",
         systemPromptRole: body,
       },
     ]);
     expect(text).toContain("### draper");
     expect(text).toContain("[source: claude]");
-    expect(text).toContain("[tier: standard]");
     expect(text).toContain("System prompt / body:");
     expect(text).toContain(body);
     expect(text).toContain("do not need read_file on plugin roots");

src/agent/agent-search.ts

@@ -68,13 +68,12 @@ function truncateAgentBody(body: string): string {
 // MAX_AGENT_SEARCH_BODY_CHARS are truncated with an ellipsis marker.
 function formatAgentProfileEntry(p: AgentProfile): string {
   const desc = (p.description ?? "").trim();
-  const tier = p.tier !== undefined ? ` [tier: ${p.tier}]` : "";
   const orch = p.orchestrator === true ? " [orchestrator]" : "";
   const source = p.source !== undefined ? ` [source: ${p.source}]` : "";
   const header =
     desc.length > 0
-      ? `### ${p.id}${tier}${orch}${source}\n${desc}`
-      : `### ${p.id}${tier}${orch}${source}`;
+      ? `### ${p.id}${orch}${source}\n${desc}`
+      : `### ${p.id}${orch}${source}`;
   const body = (p.systemPromptRole ?? "").trim();
   if (body.length === 0) return header;
   return `${header}\n\nSystem prompt / body:\n${truncateAgentBody(body)}`;

src/agent/default-agents.ts

@@ -8,7 +8,6 @@ export const defaultAgentsPlugin: AgentPlugin = {
     {
       id: "greybeard",
       description: "Seasoned architect — reviews for design, constraint ownership, and backwards compatibility",
-      tier: "clever",
       systemPromptRole:
         "You are a seasoned software architect with decades of experience. " +
         "You review code and designs for architectural soundness, constraint ownership " +
@@ -19,7 +18,6 @@ export const defaultAgentsPlugin: AgentPlugin = {
     {
       id: "critique",
       description: "Code quality reviewer — tests assumptions, finds edge cases and security smells",
-      tier: "standard",
       systemPromptRole:
         "You are a critical code reviewer focused on code quality, test coverage, " +
         "edge cases, and security-adjacent issues. You challenge assumptions, look for " +

src/agent/profile-types.ts

@@ -28,9 +28,8 @@ export type CapabilityFilter = {
   tools: string[];
 };
 
-// A single provider/model/effort combo an agent can run on. Mirrors a tier leg
-// but carries an optional reasoningEffort so an agent can pin "Sonnet + medium"
-// or "Grok + high" without going through the tier abstraction.
+// A single provider/model/effort combo an agent can run on, so an agent can
+// pin "Sonnet + medium" or "Grok + high".
 export type InferenceLeg = {
   provider: string;
   model: string;
@@ -50,12 +49,9 @@ export type AgentProfile = {
   // Unique identifier, used in workflow steps as `agent: "greybeard"`.
   id: string;
   description?: string;
-  // Provider tier alias for this agent. Resolved via settings.tiers to a
-  // concrete provider and model assignment. Used when `inference` is absent.
-  tier?: "fast" | "standard" | "clever";
-  // Explicit per-agent model selection. Takes precedence over `tier` when set,
-  // so an agent can declare "Sonnet + medium reasoning" without going through
-  // the user's tier config. See InferenceSpec for resolution rules.
+  // Explicit per-agent model selection. When absent, the agent runs on the
+  // parent session's active provider/model. See InferenceSpec for
+  // resolution rules.
   inference?: InferenceSpec;
   // Optional tool restriction. Controls which tools the sub-agent can call.
   capabilities?: CapabilityFilter;

src/agent/profiles.ts

@@ -40,7 +40,6 @@ const InferenceSpecSchema = type({
 const AgentProfileSchema = type({
   id: "string",
   "description?": "string",
-  "tier?": "'fast' | 'standard' | 'clever'",
   "inference?": InferenceSpecSchema,
   "capabilities?": CapabilityFilterSchema,
   "systemPromptRole?": "string",

src/config.test.ts

@@ -538,14 +538,11 @@ describe("buildProviderCatalog", () => {
     });
   });
 
-  test("runtimeSettingsWithCatalog overlays OAuth catalog entries for tier resolution", () => {
+  test("runtimeSettingsWithCatalog overlays OAuth catalog entries for provider resolution", () => {
     const disk = {
       providers: {
         openai: { baseURL: "https://api.openai.com/v1", apiKey: "sk", models: ["gpt-4o"] },
       },
-      tiers: {
-        clever: { provider: "xai/work", model: "grok-4" },
-      },
     };
     const catalog = [
       {
@@ -568,7 +565,6 @@ describe("buildProviderCatalog", () => {
       apiKey: "xai-token",
       models: ["grok-4"],
     });
-    expect(runtime.tiers).toEqual(disk.tiers);
     // Disk persist path still strips OAuth.
     expect(providerCatalogToSettings(catalog, "openai", disk).providers["xai/work"]).toBeUndefined();
   });
@@ -633,7 +629,6 @@ describe("buildProviderCatalog", () => {
       agentModelFallback: "none",
       shell: { timeoutMs: 30_000, maxTimeoutMs: 120_000 },
       tools: { timeoutMs: 60_000 },
-      tiers: { fast: { provider: "fp", model: "fp-large" } },
       workflowProfiles: { fast: { implement: "fp-large" } },
     };
     const settings = providerCatalogToSettings(

src/config/index.ts

@@ -47,10 +47,6 @@ import {
   normalizeOpenAICompatibleBaseURL,
   resolveProvider,
   type MCPServerConfig,
-  type ProviderTier,
-  type TierAssignment,
-  type TierConfig,
-  type TierDefinition,
   type ResolvedProvider,
   type Settings,
   type ProviderSettings,
@@ -306,9 +302,8 @@ export type Config = {
   workflow?: string;
   // Deprecated no-op retained for CLI compatibility.
   noWorkflow: boolean;
-  tiers?: Partial<Record<ProviderTier, import("./settings.js").TierConfig>>;
   /**
-   * Runtime settings view for tier/provider resolution. Includes OAuth provider
+   * Runtime settings view for provider resolution. Includes OAuth provider
    * projections from the live catalog that are never written to settings.json.
    * Do not pass this object to saveGlobalSettings — rebuild with
    * providerCatalogToSettings (or re-read disk) before any persist.
@@ -580,8 +575,7 @@ export async function loadConfig(
       : settings?.mcpServers !== undefined
         ? { mcpServers: settings.mcpServers, mcpServersSource: "global" as const }
         : { mcpServersSource: "none" as const }),
-    ...(settings?.tiers !== undefined ? { tiers: settings.tiers } : {}),
-    // Runtime view includes OAuth projections so tier resolution can see
+    // Runtime view includes OAuth projections so inference resolution can see
     // Codex/xAI providers that are never written to settings.json. Not safe
     // to persist as-is — use providerCatalogToSettings or re-read disk.
     ...(settingsForResolution !== null ? { settings: settingsForResolution } : {}),
@@ -613,7 +607,7 @@ export function catalogEntryAsProviderSettings(entry: ProviderCatalogEntry): Pro
 }
 
 // Overlay the full live catalog (including OAuth profiles) onto settings for
-// runtime tier/provider resolution. OAuth credentials live in home auth stores
+// runtime provider resolution. OAuth credentials live in home auth stores
 // and are stripped from settings.json; the catalog is the source of truth for
 // which OAuth providers are available right now. Never pass the result to a
 // disk write path — use providerCatalogToSettings for persistence.