Responsive fix, design system, SEO, brand mark, and creator page

README.md

@@ -44,6 +44,10 @@ npx tsc --noEmit   # type check
 npm run build      # static export build
 ```
 
+## Intentionally unlinked pages
+
+- `/creator` (`src/app/creator/page.tsx`) is the creator profile. It is indexable (200, self-referencing canonical, listed in `sitemap.xml`) but deliberately absent from the header, sidebar, footer, and search index, and nothing links to it internally. That is a product decision, not an oversight. Keep it that way unless the decision changes.
+
 ## Machine-readable endpoints
 
 - `/changelog.xml`: RSS feed for releases.

content/docs/core-concepts/agent-loop.md

@@ -3,6 +3,8 @@ title: The Agent Loop
 slug: agent-loop
 description: One async generator drives every run, one iteration at a time.
 order: 1
+metaTitle: "The agent loop: one generator drives every run"
+metaDescription: "How one generator runs each iteration: render the prompt, stream the model, run tool calls, record the step, and stop at a final answer or the step limit."
 ---
 
 The whole framework hangs off a single async generator, `AgentSession._run_stream`. `run()` is a thin dispatcher: in streaming mode it hands back that generator, and in blocking mode it drains the generator and returns the final answer. There is no second implementation to keep in sync.

content/docs/core-concepts/memory.md

@@ -3,6 +3,8 @@ title: Memory
 slug: memory
 description: A run is a list of typed steps that serialize, replay, and resume.
 order: 3
+metaTitle: "Memory: typed steps that persist and resume"
+metaDescription: "A run is a list of TaskStep, ActionStep, and FinalStep objects. Each renders into model messages and saves as JSON for persistence, replay, and resuming."
 ---
 
 A run is recorded as a list of typed steps rather than a flat list of messages. Steps know how to render themselves back into model messages and how to serialize to JSON, which gives you persistence, replay, and multi-turn continuation without extra machinery.
@@ -41,7 +43,7 @@ await restored.run("A follow-up", reset=False)
 
 Loading validates the payload as it rebuilds the steps. A malformed or incompatible payload raises `MemoryLoadError` rather than silently producing a broken session. See [Errors](/docs/reference/errors) for the full exception hierarchy.
 
-## What this unlocks
+## What this enables
 
 Because the full run state lives in one serializable object, several features fall out for free:
 

content/docs/core-concepts/skills.md

@@ -3,6 +3,8 @@ title: Skills
 slug: skills
 description: Instruction folders the model loads on demand, catalog first.
 order: 5
+metaTitle: "Skills: instruction folders loaded on demand"
+metaDescription: "Put a SKILL.md folder in your project. The model sees its name and summary in a catalog and loads the full instructions only when it decides it needs them."
 ---
 
 A skill is a folder containing a `SKILL.md` file: YAML frontmatter followed by a markdown body of instructions. Skills let you package expertise, and progressive disclosure keeps them cheap until the model actually needs them.

content/docs/core-concepts/streaming-events.md

@@ -3,6 +3,8 @@ title: Streaming Events
 slug: streaming-events
 description: A small set of typed events reports everything the loop does.
 order: 4
+metaTitle: "Streaming events: typed events from every run"
+metaDescription: "The small set of typed events the loop emits while it runs: text deltas, tool calls, tool results, and the final event, and how to consume them."
 ---
 
 The loop communicates progress through a small set of frozen event types. Pass `stream=True` to `run()` and you get an async iterator of these events; leave it off and the same generator is drained for you. One code path serves both modes.

content/docs/core-concepts/tools.md

@@ -3,6 +3,8 @@ title: Tools
 slug: tools
 description: Plain Python functions become schema-validated tools with @tool.
 order: 2
+metaTitle: "Tools: plain Python functions with @tool"
+metaDescription: "Turn a typed Python function into a schema-validated tool with @tool. Docstrings become descriptions, exceptions become observations, timeouts per call."
 ---
 
 A tool is any Python function wrapped with `@tool`. The function name becomes the tool name, the docstring summary becomes the description, and the type hints plus a Google-style `Args:` section become the JSON Schema the model sees.

content/docs/getting-started/installation.md

@@ -3,6 +3,8 @@ title: Installation
 slug: installation
 description: Install agentling with pip or uv and configure your provider key.
 order: 2
+metaTitle: "Install Agentling with pip or uv"
+metaDescription: "Install agentling with pip or uv on Python 3.11 or newer, set an API key for OpenAI or any compatible provider, and check the install in one command."
 ---
 
 Install agentling from PyPI:

content/docs/getting-started/introduction.md

@@ -3,6 +3,8 @@ title: Introduction
 slug: introduction
 description: What Agentling is, the idea behind it, and when to reach for it.
 order: 1
+metaTitle: "What Agentling is and when to use it"
+metaDescription: "What Agentling is, the loop it is built on, what you get (typed memory, streamed events, self-healing tools, skills), and when to use something else."
 ---
 
 Agentling is a small async Python framework for building tool-using agents. It gives you a clean ReAct loop, typed memory, streaming events, and recoverable failures in an auditable codebase.

content/docs/getting-started/quickstart.md

@@ -3,6 +3,8 @@ title: Quickstart
 slug: quickstart
 description: Build, run, and stream your first agent in a few lines of Python.
 order: 3
+metaTitle: "Quickstart: build and stream your first agent"
+metaDescription: "Define a tool with @tool, create an Agent with a model, run it, then stream events from the same loop. About 30 lines of Python from start to finish."
 ---
 
 This page takes you from an empty file to a streaming agent with one tool.

content/docs/guides/_section.md

@@ -1,7 +1,7 @@
 ---
 title: Guides
 slug: guides
-description: Task focused guides for building production ready agents.
+description: "How to run agents in production: sessions, persistence, interruption, error handling, and other providers."
 icon: compass
 order: 3
 ---

content/docs/guides/error-handling.md

@@ -3,6 +3,8 @@ title: Error Handling
 slug: error-handling
 description: Self-healing tool errors, timeouts, redaction, output truncation, and the retry behavior of the model layer.
 order: 4
+metaTitle: "Error handling: self-healing tools and retries"
+metaDescription: "How tool exceptions become observations the model can fix, per-call timeouts, output truncation, error redaction, and the retry policy of the model layer."
 ---
 
 agentling treats failure as something the model can recover from, not something that kills the run. This guide covers the layers of that design: self-healing tool errors, timeouts, redaction, and how the model layer retries.

content/docs/guides/interruption.md

@@ -3,6 +3,8 @@ title: Interruption
 slug: interruption
 description: Stop a running agent gracefully with session.interrupt() and resume it exactly where it paused.
 order: 3
+metaTitle: "Interrupt a running agent and resume it"
+metaDescription: "Stop a running agent gracefully with session.interrupt(), what the loop does at the next safe point, and how to resume from the step where it paused."
 ---
 
 Sometimes you need to stop an agent mid-run: a user hits a stop button, a deadline passes, or your process receives a shutdown signal. agentling supports graceful interruption that preserves the run so you can resume it later.

content/docs/guides/other-providers.md

@@ -3,6 +3,8 @@ title: Using Other Providers
 slug: other-providers
 description: Point OpenAIModel at any OpenAI-compatible endpoint, or implement the Model protocol for a custom provider.
 order: 5
+metaTitle: "Use other providers with OpenAIModel or Model"
+metaDescription: "Point OpenAIModel at any OpenAI-compatible endpoint, or implement the small Model protocol to plug in a provider of your own."
 ---
 
 agentling ships one provider adapter, `OpenAIModel`, but it is not tied to OpenAI. The adapter speaks to any OpenAI-compatible chat-completions endpoint, and the `Model` protocol lets you write your own adapter for anything else.

content/docs/guides/persistence-and-resuming.md

@@ -3,6 +3,8 @@ title: Persistence and Resuming
 slug: persistence-and-resuming
 description: Serialize a session's memory to JSON, restore it later, and continue runs with reset=False.
 order: 2
+metaTitle: "Persist a session to JSON and resume it later"
+metaDescription: "Serialize a session's memory to JSON, restore it in a new process, and continue the run with reset=False so the model picks up exactly where it stopped."
 ---
 
 Every session records its run as typed memory that serializes to JSON. That gives you persistence, replay, and multi-turn continuation without any extra infrastructure.

content/docs/guides/sessions-and-concurrency.md

@@ -3,6 +3,8 @@ title: Sessions and Concurrency
 slug: sessions-and-concurrency
 description: How Agent and AgentSession split configuration from run state, and how to run agents safely under concurrency.
 order: 1
+metaTitle: "Sessions and concurrency: Agent vs AgentSession"
+metaDescription: "How Agent holds configuration and AgentSession holds per-run state, and the rules for sharing agents across tasks and running sessions concurrently."
 ---
 
 agentling separates what an agent is from what a run does. An `Agent` is immutable configuration, and an `AgentSession` holds the state of one conversation. Understanding that split is the key to running agents safely under concurrency.

content/docs/reference/agent-api.md

@@ -3,6 +3,8 @@ title: Agent API
 slug: agent-api
 description: Every Agent constructor parameter, the run method, and the AgentSession surface.
 order: 1
+metaTitle: "Agent API: constructor, run(), and AgentSession"
+metaDescription: "Every Agent constructor parameter and its default, what run() and start() do, and the AgentSession surface for multi-turn, interruptible runs."
 ---
 
 `Agent` is immutable configuration: a model, some tools, and settings. Build one once, share it freely, and let sessions carry the per-run state. This page lists every knob on the constructor and the semantics of `run()` and `start()`.

content/docs/reference/errors.md

@@ -3,6 +3,8 @@ title: Errors
 slug: errors
 description: The exception hierarchy, which failures the model recovers from, and what redact_errors changes.
 order: 4
+metaTitle: "Errors reference: the exception hierarchy"
+metaDescription: "Every exception under AgentlingError, which failures the loop recovers from on its own, and what redact_errors changes about what the model sees."
 ---
 
 Every exception agentling raises lives under a single base class, `AgentlingError`, so one `except` clause can catch anything the framework produces. Just as important is what does not raise: most tool failures are converted into observations the model can recover from instead of crashing the run.

content/docs/reference/events.md

@@ -3,6 +3,8 @@ title: Events
 slug: events
 description: Every streaming event type, what it carries, and how to consume the stream.
 order: 3
+metaTitle: "Events reference: every streaming event type"
+metaDescription: "Every event the loop can emit, the fields each one carries, and how to consume the stream in blocking and streaming modes."
 ---
 
 When you call `run(task, stream=True)`, the loop communicates progress through a small set of frozen event types. This page is the reference for each one and for `print_events`, the ready-made consumer.

content/docs/reference/openai-model.md

@@ -3,6 +3,8 @@ title: OpenAIModel
 slug: openai-model
 description: The built-in provider adapter, its retry policy, and the Model protocol for custom adapters.
 order: 2
+metaTitle: "OpenAIModel reference and the Model protocol"
+metaDescription: "The built-in provider adapter: constructor options, its retry policy, how it streams, and the Model protocol a custom adapter must implement."
 ---
 
 `OpenAIModel` is an adapter for any OpenAI-compatible chat-completions endpoint. It is the only place in agentling that knows OpenAI's wire format: it converts framework-neutral `ChatMessage` lists into provider payloads on the way out and converts responses back on the way in.

content/pages/contributing.md

@@ -2,6 +2,8 @@
 title: Contributing
 slug: contributing
 description: How to set up a development environment, run the checks, and contribute to agentling.
+metaTitle: "Contributing: setup, checks, pull requests"
+metaDescription: "Set up a dev environment with uv, run the test, lint, format, and type checks that CI runs, and open a pull request against the agentling repository."
 ---
 
 Thanks for your interest in improving agentling. The framework stays deliberately small, so thoughtful contributions matter more than big ones.

content/pages/privacy.md

@@ -2,6 +2,8 @@
 title: Privacy Policy
 slug: privacy
 description: What data this website collects, what it does not, and your choices.
+metaTitle: "Privacy policy: what this site collects"
+metaDescription: "What this website collects, what it does not, which third parties are involved, and the choices you have."
 ---
 
 Last updated: 7 July 2026

content/pages/terms.md

@@ -2,6 +2,8 @@
 title: Terms of Use
 slug: terms
 description: The terms that apply to this website and to the agentling open source software.
+metaTitle: "Terms of use for this site and the software"
+metaDescription: "The terms that apply to using this website and to the agentling open source software, which is MIT licensed."
 ---
 
 Last updated: 7 July 2026

next.config.ts

@@ -12,6 +12,15 @@ const nextConfig: NextConfig = {
   outputFileTracingIncludes: {
     "/**": ["./content/**/*", "./src/assets/fonts/**"],
   },
+  // URLs that used to be linked from the site and still receive requests.
+  async redirects() {
+    return [
+      { source: "/feed.xml", destination: "/changelog.xml", permanent: true },
+      { source: "/blog", destination: "/changelog", permanent: true },
+      { source: "/blog/:path*", destination: "/changelog", permanent: true },
+      { source: "/roadmap", destination: "/changelog", permanent: true },
+    ];
+  },
 };
 
 export default nextConfig;