- Home
- Guides
All guides
Newest first. Each guide answers one operating question and says which tool versions and documentation it was checked against.
- GUIDEManchester Study Agency Fees: What to Confirm in Writing Before You SignAgent costs for Manchester applications are usually paid by the institution, though some agents charge students extra. Confirm your own position in writing.
- GUIDELlama-modchip Runs a 284B DeepSeek MoE in 64GB of RAM: The Patch, the Flags and the HeadroomWhat the patched fork, GGML types 42 and 43, AVX-512 fast paths and the memory margin actually require.
- GUIDEApplying to UTS: Which Agent Services Are Worth ConsideringBefore signing with anyone for a UTS application, check the service list, the fee schedule and where the advice stops. Here is what to file and ask.
- GUIDEQCon Shanghai Preview: How Harness Engineering Constrains SRE AgentsA preview of Wang Jie’s QCon Shanghai 2026 session on using GitOps guardrails, human review, multi-agent collaboration, and experience reuse to give SRE agents a controlled path into production.
- SETUPHow to Add MCP Servers to Claude Code and Keep Their Tools Safe to CallA Claude Code MCP guide to adding remote servers, choosing scope, checking connection status, and containing calls with server- and tool-level permission rules.
- LIBRARYOperating an Agentic RAG Agent with Citation Gates and Recovery DrillsA practical operating guide to agentic RAG that explains the template's Agno, Gemini and OpenAI structure and treats citation gates and recovery drills as external controls, not built-in features.
- LIBRARYOperating a Browser MCP Agent with Isolation, Action Limits, and Visual VerificationHow the Browser MCP Agent template runs, why its default launch does not provide isolation or action limits, and which operating controls and visual checks to add.
- FRAMEWORKClaude Agent SDK vs OpenAI Agents SDK for Tool-Building TeamsClaude Agent SDK exposes the tools, agent loop and context management that power Claude Code in Python and TypeScript, while OpenAI Agents SDK provides provider-agnostic workflows, multiple tool categories, guardrails and handoffs.
- COORDHow to Run Claude Code Agent Teams Without Overlapping Repository EditsHow to use Claude Code agent teams without overlapping repository edits by enabling the experimental mode, assigning distinct file sets, and checking task completion.
- COORDHow to Run Concurrent Coding Agents in Git Worktrees Without Merge CollisionsA Git worktree gives each coding agent a separate checkout, while path ownership and atomic shared-file edits reduce—but do not eliminate—merge collisions.
- SETUPHow to Configure Codex CLI for Repository Work: config.toml, Sandbox, Hooks, and MCPUse Codex config.toml to control repository-scoped settings, workspace network access, hook trust, and MCP servers, then verify every layer before agent edits begin.
- FRAMEWORKCrewAI vs LangGraph for Role-Based Agent Teams: Handoffs, State, and ControlA CrewAI vs LangGraph comparison shows CrewAI's explicit agent roles and delegation alongside LangGraph's shared state, state-driven handoffs, and checkpointed graph control.
- SETUPCursor vs Claude Code for Repository Work: A Practical Operator ComparisonA practical Cursor vs Claude Code comparison for repository work, including Cursor's isolated cloud agents and Claude Code's multi-surface workflow.
- SETUPHow to Design AGENTS.md Files That Carry Instructions Across Coding AgentsA practical AGENTS.md design uses root guidance, nested package files, explicit ownership, and loader-specific checks to carry one reliable instruction system across coding agents.
- COORDHow to Design Multi-Agent Systems with Handoffs, Shared State, and VerificationA practical guide to multi agent systems that gives handoffs an ownership contract, shared state an update contract, and verification an end-state check.
- COSTHow to Estimate and Control Claude Code Costs by Task TypeClaude Code pricing depends on API token consumption or included Pro and Max subscription usage, while a practical task estimate accounts for model, context, cache, output, thinking, and subagent use.
- FRAMEWORKLangChain vs LangGraph for Stateful Agent Workflows: What Changes in PracticeLangChain vs LangGraph differs at the architecture level: LangChain provides the agent harness, while LangGraph provides graph-based orchestration for stateful execution.
- VERIFYHow to Make Multi-Agent Workflows Deployable with Review GatesA practical guide to making multi-agent workflows deployable by bounding agent writes, pausing sensitive actions, and enforcing reviews before protected merges and deployments.
- COORDHow to Structure Parent, Worker, and Reviewer Agents in Claude CodeA practical claude code multi agent pattern that keeps coordination in the main conversation, delegates bounded work to separate-context subagents, and reviews changes against explicit criteria.
- SETUPHow to Set Up Playwright MCP for Repeatable Browser VerificationSet up Playwright MCP with npx, verify actions through accessibility snapshots, and isolate browser state so concurrent checks do not share one profile.
- COSTHow to Route Claude Code Through OpenRouter Without Breaking Tool CallsThe documented Claude Code router setup through OpenRouter uses the OpenRouter base URL, ANTHROPIC_AUTH_TOKEN, a blank ANTHROPIC_API_KEY, and the tools parameter on every tool-result request.
- POSTMORTEMWhy Claude Code Got Slow: A Post-Mortem of Context, Tools, and RetriesExplains why Claude Code can feel slow when context, tool output, cache resets, or transient-error retries expand the work, with documented recovery commands.
- POSTMORTEMWhy Claude Code Stopped Working: A Post-Mortem of Permissions, MCP, and Stale StateClaude Code not working usually points to merged settings, permission precedence, MCP approval or paths, authentication, limits, context, or stale session state.
- POSTMORTEMWhy Multi-Agent LLM Systems Fail: Lost State, Broken Handoffs, and False CompletionA MAST analysis of 1642 annotated traces from 7 frameworks explains why multi-agent LLM systems fail across system design, inter-agent misalignment, and task verification.
The first guides are being written and checked against each tool's own documentation.