walmsles

3 ms·

1.
by walmsles 5mo ago
The Deno sandbox is the answer here — network access is restricted to an allowlist, and the execution environment has scoped permissions. The agent builds tools within those constraints, it can't reach anything you haven't explici
2.
by walmsles 5mo ago
Not for a lack of trying! Had to enforce tool building through code as models tend to just execute arbitrary code when allowed.
3.
by walmsles 5mo ago
This is essentially ADRs — capturing what the agent learned and why. The manual trigger is the interesting constraint though; the hard part is teaching the agent to recognise the moment a decision worth recording has been made, without bein
4.
by walmsles 5mo ago
The three bootstrap tools are a partial answer to (1) — the tool surface never grows, only the registry does, so context pollution is bounded by the search interface rather than the full tool list. Whether the registry search stays useful a
5.
by walmsles 5mo ago
IKR!!!!
6.
by walmsles 5mo ago
The registry itself is searchable. The system prompts guides the agent to search it to find tools. Right now its a naive implementation as it's a local tool. I am exploring the idea of more structured policy here. It's not net
7.
by walmsles 5mo ago
It's an open experiment, the utility of tendril is the concept. I am more curious about how good can the tool making get. Frontier models tend to be very specific about what they build so don't get specific bloat (yet).
8.
by walmsles 5mo ago
I built this while working on a coding agent that kept starting cold every session. The deeper problem was that agent frameworks give you what a tool does and how to call it, but no structured answer to when — when should a tool fire autono
9.
Tendril – a self-extending agent that builds and registers its own tools(github.com)
81 points by walmsles 5mo ago|32 comments
10.
by walmsles 7mo ago
The portal is the B2B component for agent monetisation platform I am building. We are also building free and open source agent discovery - boring federated infra servicekind of like DNS - to ensure Agent discovery is free from lock-in and
11.
by walmsles 7mo ago
I built Tethyr Cloud for the AWS AIdeas competition - it's a B2B agent federation platform addressing a problem I see emerging: agent discovery is fragmenting across proprietary registries with vendor lock-in. The architecture has two
12.
Tethyr Cloud: Open Agent discovery, zero vendor lock-in (AIdeas Semi-finalist)(builder.aws.com)
1 points by walmsles 7mo ago|3 comments
13.
by walmsles 8mo ago
I built a library that uses an LLM as an orchestrator to coordinate multiple agents at runtime. You define what each agent does in markdown files using RFC 2119 constraints (MUST, SHOULD, MAY), and the orchestrator figures out who to call a
14.
Show HN: Sop-agents – Write Markdown, get coordinated agents(github.com)
1 points by walmsles 8mo ago|1 comments
15.
by walmsles 9mo ago
Hi HN, I built this because MCP servers run with full user permissions — they can see your SSH keys, AWS credentials, everything. Most people just copy configs from READMEs and hope for the best. I wanted container isolation without the Doc
16.
Show HN: Run-MCP – Run MCP servers securely in containers(github.com)
2 points by walmsles 9mo ago|1 comments