Skip to content

Topic

Agent Context Management

Making the inputs that shape an agent's next action - project, attachments, rules, model, tools, permissions, session state - explicit and inspectable.

Current stories

build1 publisher

Pi 1.0 runs MCP tools inside a sandbox so their output stays out of context

Earendil's Pi 1.0 coding agent ships native MCP support after more than a year of criticising the protocol, running tool calls inside a WASM sandbox. The team credits a more stateless July 2026 spec and a sandbox it already needed for non-LLM models, so MCP support was a small addition.

Publishers:dev.to

Reality

Evidence40
Adoption35
Hype gap+20
Incentives40
Confidence40
build1 publisher

Long agent sessions bill the same context 32 times over

A developer parsed his own Claude Code transcripts and found 1.99 billion cached reads against 62 million writes, which turns the length of your context into a recurring charge on every remaining turn rather than a setting you pick once.

Publishers:dev.to

Reality

Evidence58
Adoption16
Hype gap−10
Incentives22
Confidence47