TokenSaver
FR

Docs

TokenSaver documentation

Everything to route, govern and observe your AI agents with TokenSaver: concepts, build paths, integrations and the API reference.

TokenSaver Docs

Governed AI control plane

TokenSaver sits between your agents, apps, and LLM providers. Every call can run the same pipeline — cache, RAG, compression, PII — whether it arrives from the CLI, an OpenAI-compatible UI, n8n, or the Python SDK. Then you observe it in Flux IA and the agentic graph.

Where TokenSaver sits
  1. Your tools & apps
  2. TokenSaver (ts_…)
  3. Governed pipeline
  4. LLM providers
Product hero — console monitoring

Flagship

Agentic graph

Live cartography of pipeline, MCP, egress, and A2A runs — loops, tasks, and tool neighborhoods. Use it to explain agent behaviour to security, product, and customers.

Flagship preview — graph

Explore the agentic graph →

Who are you?

Docs are organised for three readers. Each path mixes explanations, diagrams, and reserved image slots for console captures.

I'm discovering the productCore concepts: pipeline, policies, keys, Flux IA, why governance matters.
I'm a developerCLI, SDK, Native REST, LLM egress, MCP, A2A — with code samples.
I need to plug something inWays to connect (API, MCP, CLI, egress) + market tools: Claude, Cursor, LibreChat, n8n…

Open-source starters: tokensaver-cli · tokensaver-egress — full matrix on Connect — how to plug in.

What TokenSaver is not

Not a new chat UI, not a replacement for OpenAI/Anthropic, not a vector DB you operate alone.

What it is

A control plane: one key (ts_…), policies per key, compatible HTTP + MCP + CLI, and full observability.

  1. Create a TokenSaver key Console → API keys → ts_…. Add provider keys in Settings if you use BYOK.
  2. Pick a path Connect a market tool, or Build with CLI/SDK/egress.
  3. Turn on modules Governance: Cache / RAG / Compression / PII as needed.
  4. Verify Flux IA rows + agentic graph hubs for the same period.

Also in Docs