Twigg claims its available hosted API stores conversations outside model providers, assembles model-sized context, and supports mid-conversation model switching, potentially eliminating bespoke persistence and context-management infrastructure for multi-provider applications.

state: seedheat: lowuncertainty: mediumconvergesscott: mediumagent-memory llm-apis agent-harnessesTwiggmdebeer

What is this?

Twigg’s own site presents a hosted, stateful LLM API: applications create a chat and send subsequent events, while Twigg stores conversations outside the model providers, fits context to each model’s window, routes requests across providers, and reports request costs. It advertises switching models mid-conversation and provides documentation and API access information, but the snippets do not independently demonstrate reliability or establish that it eliminates bespoke persistence and context-management infrastructure. The supplied results do not establish who operates this API or mdebeer’s role; other Twigg listings describe a branching research workspace, without clearly establishing its relationship to this API, and the similarly named Twig result is on a different domain.

Why it matters to Scott

Twigg’s advertised API packages the model-independent continuity advocated in Scott’s Retail MCP Is the Doorway Not the Memory ebook into a potential hosted alternative to Ask’s local conversation persistence, lossy compaction and LiteLLM routing; the supplied radar hits cover adjacent developments, not Twigg itself. This creates a concrete build-versus-buy evaluation, but the material does not establish faithful handoffs, attention-aware context selection or an export/recovery path that would satisfy his Sovereign Software Assurance framework, so eliminating bespoke infrastructure remains unproven.

ip:source.retail-mcp-is-the-doorway-not-the-memory-ebookdev:project.askdev:technology.litellmip:framework.context-engineeringip:framework.sovereign-software-assuranceradar:openheim-multiprovider-agent-runtimeradar:anansi-open-memory-apiradar:concept.context-managementradar:concept.model-routing

queries asked of Scott's wikis

provider-independent conversation state and model portability
agent harness persistence and context assembly
context window budgeting truncation and memory retrieval
multi-provider routing and mid-conversation model switching
build versus buy hosted agent memory infrastructure
conversation data ownership and infrastructure lock-in

Measured heat

now 0 pts/hpeak 0 pts/hcomments 0/hpeers p14momentum: steady2 platformsage 602h

points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion

How the heat travelled

09-16 14:29 (minted)	⭐ origin echo-reconstructed	Twigg describes a hosted context store, assembler, and model router: users create a chat, send subsequent events, switch models mid-conversa Twigg on blog (echo) · attributed from hn.story.49727072 · published time unknown	—
09-16 13:53	first on hacker news · published · lag ?	Show HN: A Stateful LLM API that works across providers mdebeer	—
09-16 13:53	amplified on hacker news 👑	hn.story.49727072 mdebeer	peak 1 · 0 comments · 106% of case engagement
09-16 14:21	our radar first saw it · lag ?	discovery anchor: hn.story.49727072	—

pace: p9 vs 1032 stories at the 336h mark (now 602h old) — behind addom-local-coding-harness (0.5x)

Evidence (2) — ⭐ canonical anchor

source	object	author	score	comments
🟧 hn	Show HN: A Stateful LLM API that works across providers Retrieved article excerpt Open article · Retrieved 2026-09-16T14:23:03.742245+00:00 # We manage the conversation. Twigg is a stateful API for interacting with LLMs. Simply create a chat, send events to it, and receive responses. The context is all managed and stored internally so you don't have to worry about it. A hosted context store, assembler and model router. Create a chat once, send only the next event, switch models mid-conversation. Conversations live outside the LLM providers. [Get started](https://twigg.ai/login) or [read the docs](https://twigg.ai/docs). Twigg is one API for LLM use. It stores the conversation, fits it to each model's context window, routes it to Anthropic, OpenAI, Google, xAI, Fireworks or OpenRouter, and reports what every request cost. This website is a JavaScript app, so a plain fetch of any page returns only this text. All documentation is published as markdown: - <https://twigg.ai/llms.txt>: index of the docs - <https://twigg.ai/llms-full.txt>: every docs page in one file - <https://twigg.ai/docs/quickstart.md>: quickstart - <https://twigg.ai/docs/api.md>: API reference - <https://api.twigg.ai/v1/catalogue/models.md>: model catalogue with prices API base URL: https://api.twigg.ai/api/v1, authenticated with an API key as a bearer token. OpenAPI spec: <https://api.twigg.ai/openapi.json>	mdebeer	1	0
🟧 echo.blog ⭐	Twigg describes a hosted context store, assembler, and model router: users create a chat, send subsequent events, switch models mid-conversa	Twigg	—	—

source

object

author

score

comments

🟧 hn

Show HN: A Stateful LLM API that works across providers

Retrieved article excerpt

Open article · Retrieved 2026-09-16T14:23:03.742245+00:00

# We manage the conversation.

Twigg is a stateful API for interacting with LLMs. Simply create a chat, send events to it, and receive responses. The context is all managed and stored internally so you don't have to worry about it.

A hosted context store, assembler and model router. Create a chat once, send only the next event, switch models mid-conversation. Conversations live outside the LLM providers.

[Get started](https://twigg.ai/login) or [read the docs](https://twigg.ai/docs).

Twigg is one API for LLM use. It stores the conversation, fits it to each model's context window, routes it to Anthropic, OpenAI, Google, xAI, Fireworks or OpenRouter, and reports what every request cost.

This website is a JavaScript app, so a plain fetch of any page returns only this text. All documentation is published as markdown:

- <https://twigg.ai/llms.txt>: index of the docs
- <https://twigg.ai/llms-full.txt>: every docs page in one file
- <https://twigg.ai/docs/quickstart.md>: quickstart
- <https://twigg.ai/docs/api.md>: API reference
- <https://api.twigg.ai/v1/catalogue/models.md>: model catalogue with prices

API base URL: https://api.twigg.ai/api/v1, authenticated with an API key as a bearer token. OpenAPI spec: <https://api.twigg.ai/openapi.json>

mdebeer

🟧 echo.blog ⭐

Twigg describes a hosted context store, assembler, and model router: users create a chat, send subsequent events, switch models mid-conversa

Twigg

—

Interpretation history

2026-09-16T14:33:07Z
grounded: converges/medium — Twigg’s advertised API packages the model-independent continuity advocated in Scott’s Retail MCP Is the Doorway Not the Memory ebook into a potential hosted alt
2026-09-16T14:29:39Z
case created — A documented API and concrete provider-independent state model establish a usable release distinct from the existing memory and routing cases.

Decision trace

09-30 21:54review_dormantscheduled targets exhausted or 28 quiet days
09-30 21:54drop_targetsquiet through full ladder or over cap 8
09-17 00:33groundTwigg’s advertised API packages the model-independent continuity advocated in Scott’s Retail MCP Is the Doorway Not the Memory ebook into a potential hosted alternative to Ask’s local conversation per
09-17 00:29createA documented API and concrete provider-independent state model establish a usable release distinct from the existing memory and routing cases.