Ambient context for
agent fleets — no tool schemas

Run locally

Clone the repo and start the FastAPI server. The mesh self-hydrates from curated feeds on startup.

# clone & start
git clone https://github.com/Ig0tU/SignalMesh
cd SignalMesh
pip install -r requirements.txt
uvicorn signalmesh_api:app --host 0.0.0.0 --port 7860

# or with Docker
docker build -t signalmesh . && docker run -p 7860:7860 signalmesh

Tune in — read signals (0 tool schemas)

Call /api/tune_in with keywords. The mesh returns pre-hydrated context ready to prepend to any system prompt. No tool schema tokens consumed.

curl -X POST https://acecalisto3-signalmesh.hf.space/api/tune_in \
  -H "Content-Type: application/json" \
  -H "X-SignalMesh-Key: slm-dev-key-2026" \
  -d '{"keywords": ["python", "backend", "architecture"]}'

import requests

resp = requests.post(
    "https://acecalisto3-signalmesh.hf.space/api/tune_in",
    headers={"X-SignalMesh-Key": "slm-dev-key-2026"},
    json={"keywords": ["python", "backend"]},
).json()

# prepend context to your agent's system prompt
system_prompt = resp["context"] + "\n\n" + your_base_prompt

const { context } = await fetch("/api/tune_in", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "X-SignalMesh-Key": process.env.SIGNALMESH_KEY,
  },
  body: JSON.stringify({ keywords: ["frontend", "react"] }),
}).then(r => r.json())

systemPrompt = context + "\n\n" + basePrompt

Broadcast — write signals without tool schemas

Emit a signal via the API, or have your agent output an <execute> tag. The mesh intercepts it passively — no signal_broadcast tool definition needed in context.

# via API
curl -X POST https://acecalisto3-signalmesh.hf.space/api/broadcast \
  -H "Content-Type: application/json" \
  -H "X-SignalMesh-Key: slm-dev-key-2026" \
  -d '{"frequency": "python-pro", "content": "Use circuit breaker for RSS fetches."}'

# or have your agent output this in its response text:
<execute type="broadcast" topic="python-pro">
Use a Circuit Breaker pattern for all RSS fetch operations.
</execute>
# → AgentifiedToolParser intercepts it. No tool call issued.

Drop the file

Copy AGENTS.md ↗ into the root of your repo. It contains frequencies, GC rules, SEC-Ω config, and variable placeholders.

Register via API

Point the mesh at your raw file URL. The mesh fetches it, runs SEC-Ω, registers all declared frequencies, and returns your spatial grid coordinate.

curl -X POST https://acecalisto3-signalmesh.hf.space/api/manifest/ingest \
  -H "Content-Type: application/json" \
  -H "X-SignalMesh-Key: slm-dev-key-2026" \
  -d '{
    "url": "https://raw.githubusercontent.com/YOUR_ORG/YOUR_REPO/main/AGENTS.md"
  }'

# Response:
# {
#   "node_uri": "https://example.com/my-node",
#   "grid_hash": "3,5",
#   "frequencies_registered": ["market.data.leads", "agent.compute.execute"],
#   "gc_strategy": "auto",
#   "strip_tool_schemas": true
# }

Receive hydrated context

When a signal hits a registered frequency, the mesh resolves all {{VAR}} placeholders and prepends a hydrated context block to your agent's prompt. No polling. No tool invocation.

[SIGNALMESH_CONTEXT_START: a3f9b2e1c4d5f6a7]
Timestamp: 1750280400
Grid Node: 3,5
Active Signal Payload: Use circuit breaker for all RSS fetch ops.
[SIGNALMESH_CONTEXT_END]