ContextOverflow

One corpus of thinking techniques for working with AI — through two doors. Your agent runs them; you learn them; both of you use the same names.

For you — what’s going wrong?

Describe the problem the way you’d complain about it. The same intent-classifying layer your agent uses will point you at the technique.

Describe what's going wrong

Or browse by problem: the eight categories are listed below.

For your agent — add the MCP

Endpoint: https://contextoverflow.org/mcp — free, no account, no key. Stateless: nothing you send is stored. Your agent gets the same techniques you learn here, and narrates them in the same vocabulary.

Claude Code

claude mcp add --transport http contextoverflow https://contextoverflow.org/mcp

claude.ai / Claude Desktop

Settings → Connectors → Add custom connector → paste https://contextoverflow.org/mcp. No OAuth fields needed.

Cursor

Add to ~/.cursor/mcp.json:

{ "mcpServers": { "contextoverflow": { "url": "https://contextoverflow.org/mcp" } } }

VS Code (Copilot agent mode)

Add to mcp.json:

{ "servers": { "contextoverflow": { "type": "http", "url": "https://contextoverflow.org/mcp" } } }

OpenAI Responses API

{ "type": "mcp", "server_label": "contextoverflow", "server_url": "https://contextoverflow.org/mcp" }

Anything stdio-only

npx mcp-remote https://contextoverflow.org/mcp

Five tools: list_categories · classify_intent · find_technique · get_technique · apply_technique — full details on the connect page.

The eight problems

The problem, as you experience it	Where to look
“My AI forgets everything between sessions”	Lost the thread
“I’m outsourcing my thinking and getting dumber”	Doing my thinking
“My AI states things that turn out to be false”	Confidently wrong
“My AI tells me I’m right even when I’m not”	Agrees with everything
“The AI stalls instead of acting”	Stalls instead of acting
“I ask something simple and get a wall of text”	Bloated answers
“It starts producing before it understands the task”	Starting blind
“The task is too big and it (or I) can’t hold it”	Problem too big

Every technique page follows the same path: the problem, the mechanism, how to apply it yourself, what your agent does with it, its failure modes, the evidence, and how to build it into a prompt — worked and annotated, never copy-paste.