Skip to content
Agent Algebra

About

An agent is a compositional computational process. Fixed as an effectful Mealy coalgebra with typed termination, discrete ticks and a commutative effect monad, it composes under eight operators: sequencing (;), choice (), parallel composition (), guarded recursion (μ), delayed feedback (Tr), supervision (Sup), checkpointing (Persist) and human approval (Gate_H). Each operator is proved well-defined on agents, shown to respect bisimilarity or a stated coarser relation, and given the laws it satisfies at that relation. For sixteen historically separate fields, from actor systems and process calculi to durable execution and reinforcement learning, the series gives an explicit translation of one named mechanism into a term of this algebra together with an adequacy result for that translation. No field is claimed to reduce to the algebra.

The laws are not only stated in the papers: a deterministic TypeScript reference implementation checks them against generated and hand-built agent terms, and every appendix row in the papers cites the file, the identifier and the test that establishes it.

Operators, by owning Part

SymbolMeaningPart
A : X ⇒ Yan agent of sequential type X to Y over (I, Σ)I
~bisimilarityI
≃trfinite-trace equivalenceI
stuttering-closed refinementI
A ; Bsequential composition with silent handoffII
A ⊕b B, A ⊕ B, A ⊕p Bguarded, nondeterministic, probabilistic choiceII
A ⊗ B, race(A, B)synchronous product, first-halt productII
A ▷ Brecovery: continue as B at A's failure reasonII
μX.F(X)guarded recursion, with derived loop and delayII
Trc(A)delayed feedback on port cIII
Chk, linkc, A ∥ BFIFO channel, linking, asynchronous interleavingIII
Act(A), BB, TSactor, blackboard, tuple spaceIII
Gate_H(A)human approval gateIII
Supσ, Persist_C, retrysupervisor, checkpointing, retry with backoffIV
Wf(G), Replayworkflow over a Kahn network, deterministic replayIV
πθ, V, A_π^Φparameterized policy, closed-loop value, shielded agentV

Background literature

Author

Matthew Long
The YonedaAI Collaboration, YonedaAI Research Collective
Chicago, IL
matthew@yonedaai.com · yonedaai.com