mesh

a flow-based programming language for ai agents.

what is mesh

mesh is designed from the ground up for how agents think:

• no state — no variables, no mutation. data flows through pipes.
• everything is a tool — built-in and external tools are the same kind of thing.
• errors are data — failures flow through, they don't crash the pipeline.
• parallel by default — independent operations run concurrently.
• self-describing — every mesh program documents itself.

quick start

# run a mesh file
mesh run examples/hello.mesh

# check syntax
mesh check examples/hello.mesh

# interactive repl
mesh repl

# list available tools
mesh --tools

example

# hello.mesh — fetch data, transform, output

http.get "https://api.github.com/repos/pokelabshq/council/commits"
  → json.parse
  → .[:3]
  → for each commit:
      format "{{.author.login}}: {{.commit.message}}"
  → print

syntax

# pipeline
tool_call arg1 key=value
  → next_tool
  → .field[:5]
  → print

# conditionals
check http.get "https://example.com/health"
  → if .status != 200:
      alert "down"

# parallel
parallel:
  branch a: fetch "url/a"
  branch b: fetch "url/b"
  → merge
  → print

# retry
retry 3, backoff 2s:
  http.get "https://flaky-api.example.com"
  → on_error:
      log "failed after retries"
      return {error: true}

language spec

see SPEC.md
website: mesh.pokelabs.org
full docs: docs.md for the full language specification.

built-in tools

category	tools
data	json.parse, json.stringify, format, type, string, number
http	http.get, http.post
collections	filter, map, reduce, sort, unique, flatten, take, skip, count, first, last, merge, length, keys, values
flow	print, log, return, wait, if, for, retry, parallel
system	env, shell, now, uuid, save, load

installation

git clone https://github.com/pokelabshq/mesh.git
cd mesh
python3 mesh.py --repl

license

mit — poke labs