You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Implement SEP-2567: Sessionless MCP via Explicit State Handles #244
Downstream consumer here. Flagging one result specific to SEP-2567 in case it saves someone the
same dead end.
We spent a chunk of July on custody over an explicit-state-handle wire — what re-presents a
handle, what revokes it, and what happens when the same handle lands on two instances — and wrote
it up as an exploratory negative result: the intuitive shape, where the handle rides in _meta and the server treats possession as authority, doesn't hold once the session is gone,
because possession is exactly what the session used to be. Making the handle short-lived and
single-redemption changes the ergonomics enough that it belongs in the design discussion rather
than being discovered during implementation.
Happy to share the write-up if useful to whoever picks this up. Related coordination is on #245,
where @jpurnell has mapped the foundation changes and has two stateless-transport fixes open
(#267, #268) that have been waiting on review since 2026-07-29.
Activity
bartonCUBE commentedon Sep 4, 2026
Downstream consumer here. Flagging one result specific to SEP-2567 in case it saves someone the
same dead end.
We spent a chunk of July on custody over an explicit-state-handle wire — what re-presents a
handle, what revokes it, and what happens when the same handle lands on two instances — and wrote
it up as an exploratory negative result: the intuitive shape, where the handle rides in
_metaand the server treats possession as authority, doesn't hold once the session is gone,because possession is exactly what the session used to be. Making the handle short-lived and
single-redemption changes the ergonomics enough that it belongs in the design discussion rather
than being discovered during implementation.
Happy to share the write-up if useful to whoever picks this up. Related coordination is on #245,
where @jpurnell has mapped the foundation changes and has two stateless-transport fixes open
(#267, #268) that have been waiting on review since 2026-07-29.